Skip to content

CLI Reference

Auto-generated from brain/ui_schema/cli_schema.json via sb docs cli-reference --write. Do not hand-edit.

Regenerate with sb docs cli-reference --write (or make docs-reference).

Total commands documented: 1257.

Command families

sb a2a

A2A bus and agent-card operations.

sb a2a ack

Acknowledge a message as complete.

Arguments:

  • message_id (required)

Flags:

  • --result-json — JSON object result payload
  • --json — Output JSON

sb a2a cancel-task

Cancel a local A2A task when it is still cancelable.

Arguments:

  • task_id (required)

Flags:

  • --metadata-json — Optional JSON metadata.
  • --json — Output JSON

sb a2a card-upsert

Insert or update an A2A agent card.

Arguments:

  • agent_id (required)
  • name (required)

Flags:

  • --capabilities-json — JSON array of capabilities
  • --version — Semantic version
  • --description — Agent description
  • --endpoint — Optional endpoint URI
  • --json — Output JSON

sb a2a cards

List registered A2A agent cards.

Flags:

  • --json — Output JSON

sb a2a consume

Consume the next pending message for an agent.

Arguments:

  • to_agent (required)

Flags:

  • --json — Output JSON

sb a2a dlq

Show dead-lettered A2A messages.

Flags:

  • --limit — Maximum DLQ rows
  • --json — Output JSON

sb a2a execute

Execute a queued local A2A task for a registered local agent.

Arguments:

  • task_or_message_id (required)

Flags:

  • --json — Output JSON

sb a2a get-task

Retrieve the current A2A task snapshot for a local queue task.

Arguments:

  • task_id (required)

Flags:

  • --history-length — Maximum number of task history messages to include.
  • --json — Output JSON

sb a2a list-tasks

List local queue-backed A2A tasks using protocol-style filters.

Flags:

  • --context-id — Filter by context id.
  • --status — Filter by task state.
  • --page-size — Page size.
  • --page-token — Cursor token for pagination.
  • --history-length — Maximum history messages per task.
  • --status-timestamp-after — Filter by status timestamp lower bound.
  • --include-artifacts — Include artifacts in the returned tasks.
  • --json — Output JSON

sb a2a nack

Mark a message failed and requeue or move it to DLQ.

Arguments:

  • message_id (required)

Flags:

  • --error — Failure reason
  • --json — Output JSON

sb a2a publish

Publish a message to the local A2A queue.

Arguments:

  • from_agent (required)
  • to_agent (required)
  • task (required)

Flags:

  • --payload-json — JSON object payload
  • --correlation-id — Optional correlation id
  • --ttl-seconds — Message TTL in seconds
  • --via-adk — Relay message to remote A2A agent via Google ADK.
  • --agent-card — Remote agent card URL (required with --via-adk).
  • --adk-model — Model used for ADK relay when --via-adk is enabled.
  • --json — Output JSON

sb a2a push-config-create

Create a task push notification configuration.

Arguments:

  • task_id (required)

Flags:

  • --url — Webhook URL.
  • --token — Optional client token.
  • --auth-scheme — HTTP auth scheme.
  • --auth-credentials — Auth credentials for the scheme.
  • --metadata-json — Optional JSON metadata.
  • --json — Output JSON

sb a2a push-config-delete

Delete a stored push notification configuration.

Arguments:

  • task_id (required)
  • config_id (required)

Flags:

  • --json — Output JSON

sb a2a push-config-get

Get a stored push notification configuration.

Arguments:

  • task_id (required)
  • config_id (required)

Flags:

  • --json — Output JSON

sb a2a push-configs

List stored push notification configurations for a task.

Arguments:

  • task_id (required)

Flags:

  • --page-size — Page size.
  • --page-token — Cursor token for pagination.
  • --json — Output JSON

sb a2a replay-dlq

Replay dead-lettered A2A messages back to the pending queue.

Flags:

  • --id — Replay one DLQ row id.
  • --message-id — Replay DLQ entries for source message id.
  • --from-agent — Filter DLQ rows by source agent.
  • --task — Filter DLQ rows by exact task text.
  • --moved-since — Filter DLQ rows moved at/after ISO timestamp.
  • --moved-before — Filter DLQ rows moved at/before ISO timestamp.
  • --limit — Maximum DLQ rows to replay.
  • --to-agent — Override destination agent.
  • --ttl-seconds — TTL for replayed messages.
  • --delete — Delete replayed rows from DLQ after replay.
  • --dry-run — Preview replay without enqueuing or deleting.
  • --json — Output JSON

sb a2a status

Show queue health and status counts.

Flags:

  • --json — Output JSON

sb a2a subscribe

Build the stream response sequence for a live local task subscription.

Arguments:

  • task_id (required)

Flags:

  • --json — Output JSON

sb act

Shortcut for sb tasks act.

sb actions

Governed action ledger and approval controls.

sb actions approve

Approve and execute a pending governed action when supported.

Arguments:

  • action_id (required)

Flags:

  • --approver — Approver label
  • --why — Optional rationale
  • --json — Output JSON

sb actions deny

Deny a pending governed action.

Arguments:

  • action_id (required)

Flags:

  • --approver — Approver label
  • --why — Optional rationale
  • --json — Output JSON

sb actions list

List governed actions.

Flags:

  • --status — Filter by action status
  • --type — Filter by action type (for example a2a_publish)
  • --approval-status — Filter by approval status
  • --limit — Maximum actions to show
  • --json — Output JSON

sb actions show

Show one governed action.

Arguments:

  • action_id (required)

Flags:

  • --json — Output JSON

sb activity

Owner-activity tracker — record / inspect last interaction per channel.

sb activity recent

List one record per channel, newest first.

Flags:

  • --json — Emit JSON.

sb activity show

Show the most-recent activity (overall or for one channel).

Arguments:

  • channel

Flags:

  • --json — Emit JSON.

sb activity touch

Record an owner-activity event for channel.

Flags:

  • --channel — Channel label (cli, chat, voice, discord, …).
  • --summary — Short note about this activity.
  • --user-id — User identifier.
  • --metadata — JSON object of extra metadata.

sb agent

Run YAML-defined kernel agents.

sb agent chat

Open an interactive chat REPL against the agent.

Arguments:

  • file (required)

Flags:

  • --provider — Provider name; defaults to YAML provider: or 'anthropic'.
  • --model — Override agent model.
  • --agent — Pick entry-point for multi-agent YAML.
  • --stream — Stream provider tokens as they arrive.
  • --session-id — When set, persist the conversation to .secondbrain/state/kernel_chat.db so subsequent runs resume it.
  • --with-antahkarana — Wire the antahkarana stack + bootstrap hooks.
  • --enable-autotune, --no-enable-autotune — Flip karma_autotune_bridge_enabled on the antahkarana config so regretted kernel runs feed the kernel_agent_prompt autotune lane (requires --with-antahkarana). The bridge stamps the agent's YAML path on the karma entry; the lane mutates that file's instructions: field. The evaluator defaults to a hermetic stub — set a provider on the lane's evaluator_config (or SB_KERNEL_AGENT_EVAL_PROVIDER) for real-LLM evaluation.
  • --max-turns

sb agent eval

Run an eval suite against an agent and print a scorecard.

Arguments:

  • suite (required)

Flags:

  • --agent-file, -a — Override the suite's agent: field with this YAML file.
  • --provider — Override the suite/case provider.
  • --model — Override agent model.
  • --fail-fast
  • --max-turns
  • --json

sb agent list-defs

List every agent declared in a YAML file (single- or multi-agent).

Arguments:

  • file (required)

Flags:

  • --json — JSON output

sb agent list-tools

List every registered ToolSpec id in the default registry.

Flags:

  • --only-safety — read_only|local_write|network|destructive
  • --json — JSON output

sb agent run

Run the agent once against the chosen provider and print the reply.

Arguments:

  • file (required)
  • prompt (required)

Flags:

  • --provider — Provider name (anthropic, openai, openai-responses, ollama, gemini, bedrock, cohere). Defaults to YAML provider: field or 'anthropic'.
  • --model — Override agent model
  • --agent — For multi-agent YAML, pick the entry-point agent by name (defaults to the file's entry_point: or the first agent listed).
  • --with-antahkarana — Wire the antahkarana stack + bootstrap hooks
  • --enable-autotune, --no-enable-autotune — Flip karma_autotune_bridge_enabled on the antahkarana config so regretted kernel runs feed the kernel_agent_prompt autotune lane (requires --with-antahkarana). The bridge stamps the agent's YAML path on the karma entry; the lane mutates that file's instructions: field. The evaluator defaults to a hermetic stub — set a provider on the lane's evaluator_config (or SB_KERNEL_AGENT_EVAL_PROVIDER) for real-LLM evaluation.
  • --max-turns
  • --json — JSON output

sb agent show

Render the resolved agent definition as JSON.

Arguments:

  • file (required)

sb agent validate

Parse a YAML agent definition and report any errors.

Arguments:

  • file (required)

Flags:

  • --json — JSON output

sb agent-enhancements

Agent harness enhancement utilities.

sb agent-enhancements harness-info

Show current agent harness enhancement feature flags and defaults.

Flags:

  • --json — Output JSON

sb agent-enhancements structured-output

Test structured output engine by extracting a typed response.

Flags:

  • --prompt, -p
  • --provider — Provider name (default: auto)
  • --json — Output JSON

sb agent-enhancements tool-select

Score and select the most relevant tools for a query using TF-IDF.

Arguments:

  • query (required)

Flags:

  • --max-tools, -k — Maximum tools to return
  • --web-mode — Web mode hint (on/off/auto)
  • --explain, -e — Show all tool scores
  • --json — Output JSON

sb agent-enhancements tool-validate

Validate arguments against a tool's declared schema.

Arguments:

  • tool_name (required)

Flags:

  • --args, -a — JSON arguments to validate

sb agent-identity

Operator controls for local agent identity.

sb agent-identity revoke

Revoke a registered agent identity locally.

Arguments:

  • agent (required)

Flags:

  • --reason — Revocation reason.
  • --actor — Audit actor.
  • --json — Output JSON.

sb agent-identity rotate-key

Rotate local signing keys used for agent identity claims.

Flags:

  • --a2a-card, --no-a2a-card — Also rotate the local A2A agent-card signing key.
  • --actor — Audit actor.
  • --json — Output JSON.

sb agent-identity trace

Show local event/tool-call trace rows carrying agent identity metadata.

Flags:

  • --agent — Agent id, slug, or canonical agent URN. Omit to show all identity events.
  • --trace-id — Filter by runtime trace id.
  • --session-id — Filter by session id.
  • --limit — Maximum rows per source.
  • --json — Output JSON.

sb agentic

Agentic Runtime — Planner→Executor→Critic loop.

sb agentic history

List past agentic runs stored in the runs directory.

Flags:

  • --limit, -n — Max runs to show
  • --json — Output JSON array

sb agentic plan

Generate a Plan for GOAL without executing it (always dry-run).

Arguments:

  • goal (required)

Flags:

  • --json — Output JSON

sb agentic replay

Show the full record of a past agentic run.

Arguments:

  • run_id (required)

Flags:

  • --json — Output raw JSON

sb agentic run

Run the full Planner → Executor → Critic loop for GOAL.

Arguments:

  • goal (required)

Flags:

  • --dry-run — Print plan without executing tools
  • --json — Output JSON
  • --provider — Force a specific LLM backend: anthropic, azure, bedrock, cerebras, cohere, fireworks, gemini, groq, hf, huggingface, local, mistral, nebius, ollama, openai, openrouter, parasail, perplexity, qwen, replicate, sgl, vertex, vllm, xai
  • --max-iterations — Override plan's max iterations (1-10)
  • --thinking-budget — Token budget for extended thinking (0=off)
  • --critic-tier — Critique model tier: fast (haiku) or deep (sonnet)
  • --session-id — Durable background session id (primarily for --detach runs)
  • --detach — Create a durable background agentic session instead of running inline.
  • --isolate-worktree — Run inside a fresh git worktree at .sb/worktrees/ (safe for parallel agentic runs).
  • --commit-gate — After the run, execute SessionCommitGate (smoke, progress log, commit or stash) in a bootstrapped project.

sb agents

Agent profile operations.

sb agents build

Build a new SecondBrain agent (interactive wizard or LLM auto-mode).

Flags:

  • --auto — Use LLM to auto-generate the agent from a description.
  • --description, -d — One-line description (required with --auto).
  • --name, -n — Optional PascalCase name hint (used with --auto).
  • --demo — Run a demo via the A2A bus after building.
  • --dry-run — Show the planned manifest without writing files.
  • --force — Overwrite existing agent files.
  • --json — Emit machine-readable JSON output.

sb agents build-adk

Generate an ADK-native multi-agent builder blueprint and scaffold.

Arguments:

  • objective (required)

Flags:

  • --name, -n — Optional name/slug override.
  • --model — Default ADK model.
  • --output-dir — Directory where the builder bundle will be created.
  • --force — Overwrite existing generated files.
  • --dry-run — Show planned outputs without writing files.
  • --check-runtime — Validate ADK runtime imports before generating the scaffold.
  • --json — Emit machine-readable JSON output.

sb agents instructions

Show the effective layered AGENTS instructions for a workspace.

Flags:

  • --cwd — Workspace path to inspect
  • --full — Include full instruction text
  • --json — Emit machine-readable JSON output.

sb agents list

List available agent profiles.

sb agents registry-archive

Archive (soft-delete) an agent — keeps the record but hides it from active listings.

Arguments:

  • name (required)

Flags:

  • --json — JSON output.

sb agents registry-export

Export each registered agent as .json into a directory.

Arguments:

  • output_dir (required)

Flags:

  • --tag — Only export agents with this tag.
  • --all — Include archived agents.
  • --json — Print a JSON summary instead of a table.

sb agents registry-info

Show full manifest for a registered agent.

Arguments:

  • name (required)

Flags:

  • --json — JSON output.

sb agents registry-integrity

Check agent registry for orphaned records.

sb agents registry-list

List agents in the AgentBuilder registry.

Flags:

  • --tag — Filter by tag.
  • --intent — Filter by intent.
  • --domain — Filter by domain.
  • --pattern — Filter by agent pattern.
  • --modality — Filter by modality.
  • --safety-profile — Filter by safety profile.
  • --all — Include non-active agents.
  • --json — JSON output.

sb agents registry-prompts

Show prompt pack contents for a registered agent.

Arguments:

  • name (required)

sb agents registry-restore

Restore an archived agent back to active status.

Arguments:

  • name (required)

Flags:

  • --json — JSON output.

sb agents registry-seed

Seed built-in agents into the registry (idempotent).

Flags:

  • --force — Re-seed even if already done.
  • --json — JSON output.

sb agents registry-snapshot

Build a combined AGENTS / skills / agents registry snapshot.

Flags:

  • --cwd — Workspace path to inspect
  • --snapshot — Snapshot name when persisting
  • --persist, --no-persist — Persist to registry_snapshots
  • --include-archived
  • --json — Emit machine-readable JSON output.

sb agents registry-update

Upsert an agent manifest from a YAML or JSON file.

Arguments:

  • file (required)

Flags:

  • --json — JSON output.

sb agp

Autogenesis Protocol — RSPL resources, SEPL self-evolution loop, and AutoTune bridge.

sb agp contract

Render skills.md-style contract for an entity type.

Arguments:

  • type_ (required)

Flags:

  • --out — Write to file instead of stdout

sb agp list

List every registered resource across all RSPL types.

Flags:

  • --type — Filter by entity type
  • --json — JSON output

sb agp rollback

Restore a prior version (writes a new version pointing to old content).

Arguments:

  • type_ (required)
  • name (required)
  • version (required)

Flags:

  • --yes — Skip confirmation

sb agp show

Show the latest (or a specific) record for a resource.

Arguments:

  • type_ (required)
  • name (required)

Flags:

  • --version — Specific version
  • --json — JSON output

sb agp tune

Run one AutoTune SEPL iteration and (conditionally) commit new config.

Flags:

  • --threshold — Accept threshold on quality_score
  • --json

sb agp versions

Show full version lineage for a resource.

Arguments:

  • type_ (required)
  • name (required)

Flags:

  • --json

sb aha

The sacred board: the few highest-leverage outcomes to act on now.

sb antahkarana

Antahkarana Stack — Vedic-inspired runtime responsibilities, layer status, and cognitive loop.

sb antahkarana chitta

Chitta memory stats or recall search.

Arguments:

  • query

Flags:

  • --top-k, -k — Max results
  • --json — JSON output

sb antahkarana chitta-eval

Run Chitta memory quality evals (staleness, contradictions).

Flags:

  • --min-strength — Stale threshold
  • --json — JSON output

sb antahkarana deep-think

Buddhi deep-think — Opus-4.7 extended-thinking judgments.

Flags:

  • --query, -q — Question to judge
  • --hypothesis, -H — Candidate hypothesis (repeat to add more)
  • --history — Show last N persisted judgments
  • --budget — Override thinking budget tokens for this run (0 = use config default)
  • --json — JSON output

sb antahkarana dhyana

Dharana/Dhyana focus mode — deep work sessions.

Flags:

  • --start, -s — Start focus session on topic
  • --duration, -d — Duration in minutes
  • --end, -e — End current session
  • --json — JSON output

sb antahkarana focus

Show current Manas attention focus.

Flags:

  • --json — JSON output

sb antahkarana goals

List or add Sankalpa goals.

Flags:

  • --add, -a — Add a new goal
  • --priority, -p — Goal priority
  • --json — JSON output

sb antahkarana karma

Show Karma action-outcome ledger.

Flags:

  • --lessons, -l — Show only lessons
  • --regrets, -r — Show only regrets
  • --actions, -a — Show pending next-actions
  • --limit, -n — Max entries
  • --json — JSON output

sb antahkarana karma-close

Complete or dismiss a karma next-action, closing the loop.

Arguments:

  • action_id (required)

Flags:

  • --dismiss, -d — Dismiss instead of completing

sb antahkarana narrative

Witness/Sakshi narrative — the cross-loop audit story.

Flags:

  • --compose, -c — Compose a fresh self-narrative now
  • --history, -H — Show last N narratives
  • --deterministic — Skip LLM rendering even if a provider is available
  • --json — JSON output

sb antahkarana process

Run a query through the Antahkarana runtime responsibility loop.

Arguments:

  • query (required)

Flags:

  • --priority, -p — Priority 0-1
  • --superhuman, -S — Enable extended cognitive features with a rich visual trace
  • --json — JSON output

sb antahkarana salience

Compute salience score for a text input (debug tool).

Arguments:

  • text (required)

Flags:

  • --json — JSON output

sb antahkarana self

Show or update Ahamkara self-model.

Flags:

  • --set-role — Add/update a role
  • --set-theme — Add/update a current theme
  • --set-value — Add/update a value
  • --json — JSON output

sb antahkarana status

Show Antahkarana Stack status and layer health.

Flags:

  • --json — JSON output

sb antahkarana trace

Show impulse→impression chain for a given trace.

Arguments:

  • trace_id (required)

Flags:

  • --json — JSON output

sb apple

Local Apple Mail/Calendar/Reminders integration operations.

sb apple calendar

Local Apple Calendar operations.

sb apple calendar attendees

Arguments:

  • identifier (required)

Flags:

  • --json

sb apple calendar get

Arguments:

  • identifier (required)

Flags:

  • --json

sb apple calendar ingest-next

Flags:

  • --project
  • --tag — Tags (repeatable)
  • --defer — Persist capture immediately and process later
  • --json

sb apple calendar next

Flags:

  • --json

sb apple calendar upcoming

Flags:

  • --start — ISO date/time
  • --end — ISO date/time
  • --calendar — Calendar name filter; repeat for multiple
  • --limit
  • --json

sb apple health

Flags:

  • --json

sb apple index

Optional local Apple mail index operations.

sb apple index rebuild

Flags:

  • --days — Only index messages received within this many days
  • --account
  • --mailbox
  • --limit-per-mailbox
  • --timeout-seconds — Per Apple Mail query timeout for index rebuild only.
  • --json

sb apple index status

Flags:

  • --json

sb apple mail

Local Apple Mail operations.

sb apple mail get

Arguments:

  • identifier (required)

Flags:

  • --live — Query Apple Mail directly instead of using the local mail index.
  • --json

sb apple mail ingest

Arguments:

  • identifier (required)

Flags:

  • --project
  • --tag — Tags (repeatable)
  • --defer — Persist capture immediately and process later
  • --live — Query Apple Mail directly instead of using the local mail index.
  • --json

sb apple mail ingest-flagged

Flags:

  • --days — Only include flagged mail from this many recent days
  • --account
  • --mailbox
  • --limit
  • --project
  • --tag — Tags (repeatable)
  • --defer — Persist captures immediately and process later
  • --live — Query Apple Mail directly instead of using the local mail index.
  • --json

Arguments:

  • query (required)

Flags:

  • --account
  • --mailbox
  • --unread-only
  • --flagged-only
  • --sender
  • --subject
  • --date-from — ISO date/time
  • --date-to — ISO date/time
  • --limit
  • --project
  • --tag — Tags (repeatable)
  • --defer — Persist captures immediately and process later
  • --live — Query Apple Mail directly instead of using the local mail index.
  • --json

sb apple mail mailboxes

Flags:

  • --account
  • --json

Arguments:

  • query (required)

Flags:

  • --account
  • --mailbox
  • --unread-only
  • --flagged-only
  • --sender
  • --subject
  • --date-from — ISO date/time
  • --date-to — ISO date/time
  • --limit
  • --live — Query Apple Mail directly instead of using the local mail index.
  • --json

sb apple reminders

Local Apple Reminders operations.

sb apple reminders completed

Flags:

  • --list — Reminder list name filter; repeat for multiple
  • --limit
  • --json

sb apple reminders get

Arguments:

  • identifier (required)

Flags:

  • --json

sb apple reminders ingest

Arguments:

  • identifier (required)

Flags:

  • --project
  • --tag — Tags (repeatable)
  • --defer — Persist capture immediately and process later
  • --json

sb apple reminders list

Flags:

  • --list — Reminder list name filter; repeat for multiple
  • --due-before — ISO date/time
  • --limit
  • --json

sb apple workflow

Cross-workflow Apple helpers.

sb apple workflow day-summary

Flags:

  • --json

sb apple workflow ingest-next-meeting-context

Flags:

  • --project
  • --tag — Tags (repeatable)
  • --defer — Persist capture immediately and process later
  • --json

sb apple workflow next-meeting-unread-mail

Flags:

  • --json

sb apple workflow recent-attendee-mail

Flags:

  • --days
  • --json

sb approvals

Cross-surface approval audit ledger.

sb approvals ledger

Show normalized approval records across chat, gateway, MCP, and sessions.

Flags:

  • --status — Filter approval request status
  • --limit — Rows to return
  • --json — Output JSON

sb approve

Shortcut for sb tasks approve.

sb asana

Asana — read tasks and projects.

sb asana auth

Save your Asana personal access token to keys.env.

sb asana ingest

Ingest Asana tasks into SecondBrain memory.

Flags:

  • --workspace — Workspace GID
  • --limit, -n

sb asana projects

List Asana projects.

Flags:

  • --workspace — Workspace GID
  • --limit, -n
  • --json

sb asana tasks

List my Asana tasks.

Flags:

  • --workspace — Workspace GID (auto-detect if blank)
  • --completed — Include completed tasks
  • --limit, -n
  • --json

sb asana workspaces

List Asana workspaces.

sb ask

Answer a question grounded in vault content with citations.

sb autodata

AutoData — agentic self-instruct that creates high-quality QA datasets from grounded sources.

sb autodata diversity

Dataset-level analysis of an AutoData run.

Arguments:

  • run_dir (required)

Flags:

  • --dedup-threshold — Token-Jaccard threshold for clustering near-duplicate questions.
  • --json

sb autodata fixture-init

Bootstrap an autotune fixture from real markdown files.

Arguments:

  • globs (required)

Flags:

  • --out — Where to write the fixture YAML.
  • --name — Fixture name (defaults to autodata.).
  • --weak-score — Default judge score for weak-solver responses on each source.
  • --strong-score — Default judge score for strong-solver responses on each source.
  • --root — Root for glob resolution (defaults to the SecondBrain repo root).
  • --json

sb autodata generate

Generate AutoData QA items from one or more grounded source files.

Arguments:

  • sources (required)

Flags:

  • --out — Output dir (default: state/autodata/).
  • --max-rounds — Inner-loop rounds per source.
  • --weak-provider — Provider for weak solver.
  • --weak-model
  • --strong-provider — Provider for strong solver.
  • --strong-model
  • --challenger-provider
  • --challenger-model
  • --judge-provider
  • --judge-model
  • --weak-n — Weak solver invocations.
  • --strong-n — Strong solver invocations.
  • --feed-karma — Record each round as a KarmaEntry (regret on rejection) so the autotune autodata lane picks up gate failures.
  • --json — Emit JSON summary.

sb autodata meta-optimize

Run the AutoData meta-optimizer over harness variants.

Flags:

  • --train — Training source files (used by mutation evaluator).
  • --validate — Validation source files (used to score variants).
  • --iterations — Number of mutate-evaluate rounds.
  • --temperature — Boltzmann sampling temperature.
  • --seed
  • --max-rounds — Per-source budget.
  • --json

sb autodata profile

Profile an AutoData run via brain.datasets.profile_dataset.

Arguments:

  • run_dir (required)

Flags:

  • --dedup — Drop near-duplicates before profiling.
  • --dedup-threshold
  • --json

sb autodata propose-validate

Meta-optimizer proposes; autotune lane judges; commit only if accepted.

Flags:

  • --pack — Autotune fixture pack.
  • --iterations — Meta-opt iterations.
  • --seed — Meta-opt RNG seed.
  • --temperature
  • --apply — Write tuning.yaml and create an autotune-style commit if the lane judges accept. Without --apply, prints the verdict only.
  • --json

sb autodata status

Show summary for an existing AutoData run.

Arguments:

  • run_dir (required)

sb autodata to-dataset

Export accepted AutoData items as a flat dataset (CSV/JSONL/JSON).

Arguments:

  • run_dir (required)

Flags:

  • --out — Output file (.csv | .jsonl | .json).
  • --format — Force format (csv|jsonl|json); inferred from suffix otherwise.
  • --dedup — Drop near-duplicate questions before exporting.
  • --dedup-threshold — Token-Jaccard threshold for dedup.
  • --json

sb automations

Automation operations over integration-backed jobs.

sb automations cleanup

Flags:

  • --older-than-days — Delete runs older than N days (uses ended_at/started_at timestamp).
  • --status — Optional run status filter.
  • --integration — Optional integration id/name filter.
  • --orphan-runs-only — Only remove runs whose integration record no longer exists.
  • --include-running — Include currently running rows in cleanup candidates.
  • --limit — Maximum rows to delete in one invocation.
  • --dry-run — Preview candidates without deleting.
  • --json

sb automations doctor

Flags:

  • --type
  • --limit
  • --max-stale-24h
  • --max-failing
  • --json

sb automations failures

Flags:

  • --since — Optional ISO timestamp floor for failed runs.
  • --type
  • --per-integration-limit
  • --limit
  • --json

sb automations fix

Flags:

  • --type
  • --limit
  • --only — Filter remediation targets: failing|stale.
  • --apply — Execute remediation actions (default is preview only).
  • --write-sync — When used with --apply, run sync actions with write mode (default is dry-run sync).
  • --force — Pass through force flag to sync/test operations.
  • --skip-tests — Do not run test actions even if planned.
  • --skip-runs — Do not run sync actions even if planned.
  • --max-actions — Maximum number of planned actions to execute.
  • --stop-on-error — Stop execution at first failed action.
  • --retries — Retry failed actions up to this many additional attempts.
  • --yes — Required with --apply --write-sync.
  • --json

sb automations fixes

Flags:

  • --limit — Maximum fix events to return.
  • --status — Optional event status filter.
  • --json

sb automations health

Flags:

  • --type
  • --limit
  • --max-stale-24h — Fail if stale_24h_count exceeds this value.
  • --max-failing — Fail if failing_count exceeds this value.
  • --json

sb automations list

Flags:

  • --type
  • --status
  • --limit
  • --json

sb automations pause

Arguments:

  • identifier

Flags:

  • --all — Pause all matching automations.
  • --type — Optional type filter for --all.
  • --status — Optional status filter for --all.
  • --dry-run — Preview pause operations without mutating.
  • --yes — Required with --all unless --dry-run.
  • --json

sb automations remove

Arguments:

  • identifier

Flags:

  • --all — Remove all matching automations.
  • --type — Optional type filter for --all.
  • --status — Optional status filter for --all.
  • --delete-runs — Also delete historical run records for this automation.
  • --dry-run — Preview removals without deleting.
  • --yes — Required with --all unless --dry-run.
  • --json

sb automations resume

Arguments:

  • identifier

Flags:

  • --all — Resume all matching automations.
  • --type — Optional type filter for --all.
  • --status — Optional status filter for --all.
  • --dry-run — Preview resume operations without mutating.
  • --yes — Required with --all unless --dry-run.
  • --json

sb automations run

Arguments:

  • identifier

Flags:

  • --due-only, --all — Run only due integrations when no identifier.
  • --type
  • --limit
  • --dry-run
  • --force
  • --json

sb automations runs

Arguments:

  • identifier (required)

Flags:

  • --limit
  • --json

sb automations test

Arguments:

  • identifier

Flags:

  • --all — Test all matching automations.
  • --type — Optional type filter for --all.
  • --status — Optional status filter for --all.
  • --limit — Maximum automations to test when using --all.
  • --force — Pass through force flag to dry-run sync probe.
  • --skip-sync — Only run config/source checks (no sync probe).
  • --json

sb autotune

Constrained autotuning for prompts, retrieval, routers, and planners.

sb autotune bench

Run the evaluator only without proposing any mutation.

Arguments:

  • lane (required)

Flags:

  • --pack — Benchmark pack
  • --json — Output JSON

sb autotune benchmark

Benchmark growth and curation.

sb autotune benchmark duplicate

Mark one benchmark candidate as a duplicate of another.

Arguments:

  • candidate_id (required)

Flags:

  • --of — Canonical duplicate target
  • --notes
  • --by
  • --json — Output JSON

sb autotune benchmark enrich

Enrich a benchmark candidate with operator-supplied expectations.

Arguments:

  • candidate_id (required)

Flags:

  • --expected-property
  • --forbidden-property
  • --expected-tool
  • --allowed-tool
  • --forbidden-tool
  • --expected-output
  • --expected-ref
  • --expected-route
  • --output-schema
  • --difficulty-score
  • --pack-suggestion
  • --notes
  • --by
  • --json — Output JSON

sb autotune benchmark ingest

Harvest benchmark candidates from a prior run artifact.

Arguments:

  • lane (required)

Flags:

  • --run-id — Source autotune run id
  • --json — Output JSON

sb autotune benchmark ingest-from-events

Harvest benchmark candidates from persisted canonical event rows.

Arguments:

  • lane (required)

Flags:

  • --session-id — Specific session id
  • --limit — How many recent events to inspect
  • --json — Output JSON

sb autotune benchmark ingest-from-repl

Harvest benchmark candidates from persisted chat transcripts and tool traces.

Arguments:

  • lane (required)

Flags:

  • --session-id — Specific chat session id
  • --limit — How many recent sessions to inspect
  • --json — Output JSON

sb autotune benchmark karma-enrich

Heuristically enrich karma-derived BenchmarkCandidates.

Flags:

  • --lane — Lane to enrich (omit for all lanes)
  • --candidate-id — Enrich one specific karma candidate; overrides --lane
  • --limit — Max candidates to enrich in one pass
  • --json — Output JSON

sb autotune benchmark list

List promoted benchmark cases for a lane.

Arguments:

  • lane (required)

Flags:

  • --json — Output JSON

sb autotune benchmark promote

Promote a triaged benchmark candidate into an active benchmark case.

Arguments:

  • case_id (required)

Flags:

  • --pack — Target pack
  • --force — Promote despite incomplete harness-grade validation; records audit metadata.
  • --by — Reviewer or automation id
  • --json — Output JSON

sb autotune benchmark reject

Reject a benchmark candidate during human triage.

Arguments:

  • candidate_id (required)

Flags:

  • --notes
  • --by
  • --json — Output JSON

sb autotune benchmark report

Report benchmark queue and active pack counts.

Arguments:

  • lane (required)

Flags:

  • --json — Output JSON

sb autotune benchmark show

Show one benchmark candidate including provenance.

Arguments:

  • candidate_id (required)

Flags:

  • --json — Output JSON

sb autotune benchmark triage

Bulk-triage proposed candidates for one lane.

Arguments:

  • lane (required)

Flags:

  • --status — New status
  • --source — Optional source_type filter
  • --json — Output JSON

sb autotune benchmark unresolved

List benchmark pressure sources not closed by self-improvement.

Flags:

  • --lane — Optional lane filter
  • --limit — Candidates to return
  • --json — Output JSON

sb autotune cycle

Self-improvement cycle records.

sb autotune cycle show

Show one self-improvement cycle with target-bound lineage.

Arguments:

  • cycle_id (required)

Flags:

  • --json — Output JSON

sb autotune diff

Print the stored diff for a prior autotune attempt.

Arguments:

  • run_id (required)

sb autotune experience

Filesystem-backed candidate experience for Meta-Harness-style search.

sb autotune experience context

Render the proposer context assembled from full-history artifacts.

Arguments:

  • lane (required)

Flags:

  • --limit — Maximum records per group
  • --json — Output JSON

sb autotune experience frontier

Show the accuracy/cost Pareto frontier for one lane.

Arguments:

  • lane (required)

Flags:

  • --limit — Maximum records to show
  • --json — Output JSON

sb autotune experience list

List full-history candidate experience artifacts for one lane.

Arguments:

  • lane (required)

Flags:

  • --limit — Maximum records to show
  • --query — Filter by text
  • --json — Output JSON

sb autotune experience show

Show one candidate experience manifest and artifact paths.

Arguments:

  • lane (required)
  • run_id (required)

Flags:

  • --json — Output JSON

sb autotune export

Write a dashboard JSON snapshot under the autotune artifacts root.

Flags:

  • --json — Output JSON file path

sb autotune health

Show lane health and circuit-breaker state.

Flags:

  • --json — Output JSON

sb autotune idea

Legacy compatibility surface for research idea memory.

sb autotune idea add

Add a research idea to idea memory (legacy compatibility surface).

Arguments:

  • lane (required)
  • title (required)

Flags:

  • --hypothesis — What should improve
  • --expected-effect — Expected effect
  • --mutation-strategy
  • --target-file — Target file; repeatable
  • --source — Idea source
  • --json — Output JSON

sb autotune idea list

List remembered research ideas (legacy compatibility surface).

Flags:

  • --lane — Lane filter
  • --json — Output JSON

sb autotune idea promote

Mark an idea as ready for sampling (legacy compatibility surface).

Arguments:

  • idea_id (required)

sb autotune idea reject

Reject an idea after review (legacy compatibility surface).

Arguments:

  • idea_id (required)

sb autotune idea show

Show one idea record (legacy compatibility surface).

Arguments:

  • idea_id (required)

Flags:

  • --json — Output JSON

sb autotune improve

Run one full self-improvement cycle.

Flags:

  • --lane — Pin a specific lane instead of auto-selecting
  • --max-lanes — Maximum distinct lanes to run
  • --pack — Benchmark pack for autotune runs
  • --ignore-pause — Allow an operator-driven cycle to bypass temporary lane pause state
  • --dry-run — Diagnose and seed ideas but do not execute runs
  • --explain-plan — Show target-bound packs and replay obligations without mutating state
  • --json — Output JSON

sb autotune lane

Lane-oriented reports and status.

sb autotune lane report

Alias for sb autotune report.

Arguments:

  • lane (required)

Flags:

  • --last — How many recent attempts to include
  • --json — Output JSON

sb autotune list

Show all available autotune lane specs.

Flags:

  • --json — Output JSON

sb autotune meta

Show cross-lane meta evaluation summary.

Flags:

  • --json — Output JSON

sb autotune meta-learn

Extract mutation strategy priors from run history.

Flags:

  • --write-chitta, --no-chitta — Store learned priors as Chitta procedural memories
  • --json — Output JSON

sb autotune promotion

Legacy compatibility surface for human-gated promotion bundles.

sb autotune promotion approve

Approve a promotion bundle for human handoff.

Arguments:

  • promotion_id (required)

Flags:

  • --notes
  • --by
  • --json — Output JSON

sb autotune promotion defer

Defer a promotion bundle for later review.

Arguments:

  • promotion_id (required)

Flags:

  • --notes
  • --by
  • --json — Output JSON

sb autotune promotion list

List promotion bundles awaiting human review.

Flags:

  • --json — Output JSON

sb autotune promotion reject

Reject a promotion bundle.

Arguments:

  • promotion_id (required)

Flags:

  • --notes
  • --by
  • --json — Output JSON

sb autotune promotion show

Show one promotion bundle.

Arguments:

  • promotion_id (required)

Flags:

  • --json — Output JSON

sb autotune report

Show recent history and aggregate report for one lane.

Arguments:

  • lane (required)

Flags:

  • --last — How many recent attempts to include
  • --json — Output JSON

sb autotune review

Legacy compatibility surface for static local review/control-plane artifacts.

sb autotune review build

Build static local review artifacts for promotions, benchmarks, and lane health.

Flags:

  • --json — Output JSON

sb autotune review open

Print the local review index path.

Flags:

  • --json — Output JSON

sb autotune run

Run the constrained optimization loop for one lane.

Arguments:

  • lane (required)

Flags:

  • --attempts — Override max attempts
  • --dry-run — Avoid git mutation; use a temp copy instead
  • --resume — Start from last accepted commit for this lane
  • --idea — Optional hint to bias candidate ordering
  • --parallel — Parallel workers
  • --pack — Benchmark pack to run
  • --session — Attach runs to a research session
  • --ignore-pause — Allow an operator-driven run to bypass temporary lane pause state
  • --json — Output JSON

sb autotune self-diagnose

Scan all signal sources and list ranked self-improvement targets.

Flags:

  • --json — Output JSON

sb autotune session

Legacy compatibility surface for named research sessions.

sb autotune session create

Create a named research session (legacy compatibility surface).

Arguments:

  • title (required)

Flags:

  • --lane — Lane to include; repeatable
  • --objective — Research objective
  • --scheduler — Scheduler name
  • --max-duration-sec
  • --max-total-runs
  • --pack — Initial benchmark pack
  • --max-parallel
  • --json — Output JSON

sb autotune session list

List recent research sessions (legacy compatibility surface).

Flags:

  • --json — Output JSON

sb autotune session report

Report status for one research session (legacy compatibility surface).

Arguments:

  • session_id (required)

Flags:

  • --json — Output JSON

sb autotune session run

Run a named research session until its budget is exhausted (legacy compatibility surface).

Arguments:

  • session_id (required)

Flags:

  • --dry-run — Use temp copies instead of git worktrees
  • --json — Output JSON

sb autotune status

Show machine-friendly status across lanes, sessions, and promotions.

Flags:

  • --json — Output JSON

sb autotune validate

Validate lane wiring, paths, and git safety preconditions.

Arguments:

  • lane (required)

Flags:

  • --json — Output JSON

sb backup

Create a timestamped backup of SecondBrain state (SQLite + vectors + artifacts).

sb bench

Latency benchmarking for providers and retrieval.

sb bench provider

Benchmark LLM provider latency with N calls.

Flags:

  • --prompt — Test prompt
  • --turns, -n — Number of calls
  • --provider — Provider name
  • --model — Model name
  • --json — Output JSON

sb bench retrieval

Benchmark retrieval pipeline latency.

Flags:

  • --query — Test query
  • --turns, -n — Number of retrievals
  • --top-k — Top-k chunks
  • --json — Output JSON

sb boot

Bootstrap graph — run and inspect startup stages.

sb boot run

Run bootstrap graph and show stage results.

Flags:

  • --skip-slow — Skip provider/MCP/plugin checks
  • --json — Output as JSON

sb brain

Intelligence subsystem operations.

sb brain chief

Proactive chief-of-staff operations.

sb brain chief list

List persisted chief-of-staff notes.

Flags:

  • --status — Filter by status
  • --limit — Max notes
  • --json — Emit JSON output

sb brain chief run

Generate proactive chief-of-staff notes.

Flags:

  • --dry-run — Preview notes without persisting
  • --limit — Maximum generated notes
  • --json — Emit JSON output

sb brain dream

Run a dream cycle to discover cross-session insights.

Flags:

  • --lookback-days — How many days to scan (0 = use config default)
  • --dry-run — Preview without persisting
  • --instructions — Free-text steering for the cycle (e.g. "focus on coding-style preferences; ignore one-off debugging notes")
  • --json — Emit JSON output

sb brain evolve

Derive strategy adjustments from unapplied feedback.

Flags:

  • --dry-run — Preview without applying
  • --json — Emit JSON output

sb brain feedback

Record user feedback for strategy improvement.

Arguments:

  • context (required)
  • correction (required)

Flags:

  • --type — correction|rating|tip
  • --session-id — Associated session
  • --json — Emit JSON output

sb brain insights

List discovered insights.

Flags:

  • --status — Filter by status (active|pending|deprecated)
  • --type — Filter by type (theme|pattern|gap|connection)
  • --query — Search insight text/title relevance
  • --limit — Max results
  • --json — Emit JSON output

sb brain reorganize

Non-destructive LTM rebuild proposals (see docs/how-to/auto-dreaming.md).

Flags:

  • --lookback-days — Recorded on the proposal for audit; reorganizer scans active LTMs irrespective.
  • --instructions — Free-text steering recorded on the proposal (mirrors Anthropic Dreams' field).
  • --dry-run — Preview the proposal without saving it.
  • --threshold — Cosine similarity threshold for merge candidates (default 0.92).
  • --json — Emit JSON output

sb brain reorganize accept

Apply a pending proposal's mutations atomically.

Arguments:

  • proposal_id (required)

Flags:

  • --json — Emit JSON output

sb brain reorganize list

List reorganization proposals.

Flags:

  • --status — Filter: pending|accepted|rejected
  • --limit — Max proposals
  • --json — Emit JSON output

sb brain reorganize reject

Mark a pending proposal as rejected.

Arguments:

  • proposal_id (required)

Flags:

  • --reason — Optional rejection reason
  • --json — Emit JSON output

sb brain reorganize show

Show a proposal with its full mutation list.

Arguments:

  • proposal_id (required)

Flags:

  • --json — Emit JSON output

sb brain research-queue

Autonomous research queue operations.

sb brain research-queue enqueue

Enqueue a research task manually.

Arguments:

  • topic (required)

Flags:

  • --query — Optional explicit research query
  • --trigger-type — Trigger source label
  • --trigger-text — Trigger context text
  • --priority — Priority (lower is higher priority)
  • --session-id — Optional session context
  • --json — Emit JSON output

sb brain research-queue list

List autonomous research queue items.

Flags:

  • --status — Filter by status
  • --limit — Max queue items
  • --json — Emit JSON output

sb brain research-queue run

Process due autonomous research queue items.

Flags:

  • --max-items — Maximum queue items to process
  • --dry-run — Preview due items
  • --json — Emit JSON output

sb bridge

File-based multi-channel task bridge — submit, inspect, complete tasks across channels.

sb bridge claim

Move a pending task into the in_progress queue (single-worker dispatch).

Arguments:

  • task_id (required)

sb bridge complete

Write a result for task_id and archive the task.

Arguments:

  • task_id (required)

Flags:

  • --body — Result body. Reads stdin if omitted. Supports marker prefixes ([deduped:...], [no-send], [REPLIED], [file:...]).
  • --status — completed | failed | deduped | skipped.
  • --error — Error message (if failed).

sb bridge handlers

List handler names registered in the default registry.

Flags:

  • --json — Emit JSON.

sb bridge list

List pending tasks (and optionally completed results).

Flags:

  • --json — Emit JSON.
  • --with-results — Include completed results.

sb bridge markers

Describe the result-body protocol markers.

sb bridge priorities

List available priorities and their source defaults.

sb bridge purge-archive

Delete archived task files older than the cutoff.

Flags:

  • --older-than-hours — Delete archived tasks older than N hours.

sb bridge show

Show one task envelope + its result if available.

Arguments:

  • task_id (required)

Flags:

  • --json — Emit JSON.

sb bridge submit

Submit a new task into the bridge inbox.

Arguments:

  • task (required)

Flags:

  • --source — Channel that originated the task.
  • --priority — urgent | normal | low (default: per-source).
  • --channel-id — Channel identifier.
  • --user-id — User identifier.
  • --tier — owner | team | other.
  • --metadata — JSON object of extra metadata.

sb bridge tick

Synchronously claim + dispatch pending tasks once (dev/debug).

Flags:

  • --json — Emit JSON.
  • --batch — Max tasks claimed in this tick.

sb bridge wait

Block until task_id has a result on file (or timeout).

Arguments:

  • task_id (required)

Flags:

  • --timeout — Max seconds to wait.
  • --poll — Seconds between polls.
  • --json — Emit JSON.

sb brief

Ask My Life: Antahkarana-powered briefing for any moment.

sb cache

Manage the LLM response cache.

sb cache clear

Clear cached responses.

Flags:

  • --expired-only — Remove only expired entries.

sb cache status

Show cache statistics.

sb capabilities

Unified capability bundles and inventory.

sb capabilities list

List workflows, agents, plugins, and plugin capabilities in one inventory.

Flags:

  • --kind — Filter by capability kind.
  • --tag — Filter by tag/category.
  • --all — Include disabled workflows and non-active agents.
  • --json — Output JSON.

sb capabilities schema

Print the canonical CapabilityBundle JSON schema.

Flags:

  • --json, --no-json — Output JSON schema.

sb capabilities show

Show one inventory capability by id.

Arguments:

  • capability_id (required)

Flags:

  • --kind — Restrict lookup by kind.
  • --all — Include disabled workflows and non-active agents.
  • --json — Output JSON.

sb capabilities validate

Validate a capability bundle without installing or running anything.

Arguments:

  • path (required)

Flags:

  • --check-paths — Resolve local file references and require them to exist.
  • --json — Output JSON.

sb capture

Capture management and status.

sb capture clip

Canonical clipboard capture path.

Flags:

  • --title, -t — Title for the clipped content
  • --tag — Tags (repeatable)
  • --project, -p — Project name
  • --defer — Persist now and process later
  • --json — Output JSON

sb capture file

Canonical file capture path for docs, notes, and transcripts.

Arguments:

  • path (required)

Flags:

  • --title, -t — Optional title override
  • --tag — Tags (repeatable)
  • --project, -p — Project name
  • --meeting — Treat file as meeting capture
  • --decision — Treat file as decision input
  • --task — Treat file as task input
  • --defer — Persist now and process later
  • --json — Output JSON

sb capture meeting

Canonical meeting capture path.

Flags:

  • --title, -t — Meeting title
  • --file, -f — Path to transcript or notes file
  • --text — Inline transcript text
  • --project, -p — Project name
  • --people — Comma-separated participant names
  • --tag — Tags (repeatable)
  • --date — Meeting date (YYYY-MM-DD)
  • --defer — Persist now and process later
  • --json — Output JSON

sb capture note

Canonical note capture path.

Arguments:

  • text

Flags:

  • --stdin — Read from stdin
  • --title, -t — Note title
  • --tag — Tags (repeatable)
  • --project, -p — Project name
  • --meeting — Mark as meeting note
  • --decision — Mark as decision
  • --task — Mark as task
  • --file, -f — Read text from file
  • --defer — Persist now and process later
  • --json — Output JSON

sb capture process

Process deferred captures that were only persisted.

Flags:

  • --limit, -n — Maximum queued captures to process
  • --json — Output JSON

sb capture quarantine-ack

Acknowledge a quarantined manifest so it stops appearing in attention views.

Arguments:

  • manifest_path (required)

Flags:

  • --note — Optional acknowledgement note
  • --json — Output JSON

sb capture quarantine-list

List quarantined watch imports from sidecar manifests.

Arguments:

  • path (required)

Flags:

  • --limit, -n — Maximum manifest records to show
  • --json — Output JSON

sb capture quarantine-report

Summarize quarantine/archive manifest outcomes for operator review.

Arguments:

  • path (required)

Flags:

  • --limit, -n — Maximum manifest records to summarize
  • --json — Output JSON

sb capture quarantine-retry

Retry quarantined watch imports from their sidecar manifests.

Arguments:

  • path (required)

Flags:

  • --limit, -n — Maximum quarantined files to retry
  • --archive, --no-archive — Move successful retries to an archive folder
  • --archive-dir — Archive folder for successful retries
  • --defer — Persist retried captures now and process them later
  • --json — Output JSON

sb capture quarantine-stale

List quarantined manifests that need operator attention.

Arguments:

  • path (required)

Flags:

  • --older-than-hours — Flag quarantined items older than this age
  • --min-retries — Flag items retried at least this many times
  • --include-acknowledged — Include already acknowledged manifests
  • --limit, -n — Maximum stale items to show
  • --json — Output JSON

sb capture quarantine-unack

Remove acknowledgement from a quarantined manifest.

Arguments:

  • manifest_path (required)

Flags:

  • --json — Output JSON

sb capture retry

Re-process captures through the pipeline.

Flags:

  • --failed, --all — Retry only failed captures (--all retries failed + persisted + classified)
  • --json — Output JSON

sb capture status

Show recent capture status and summary counts.

Flags:

  • --limit, -n — Number of recent captures
  • --json — Output JSON

sb capture web

Capture a URL and operator note into the unified capture pipeline.

Arguments:

  • url (required)

Flags:

  • --title, -t — Optional title override
  • --note — Short operator note
  • --tag — Tags (repeatable)
  • --project, -p — Project name
  • --defer — Persist now and process later
  • --json — Output JSON

sb capture worker

Run a background worker for deferred capture normalization.

Flags:

  • --once — Run one queue-processing cycle and exit
  • --batch-size — Maximum queued captures to process per cycle
  • --interval-seconds — Sleep interval between queue-processing cycles
  • --json — Output JSON

sb chat

Start interactive chat REPL with local retrieval and optional web search.

sb chat-doctor

Report sb chat reliability diagnostics from the turn journal.

sb checkpoint

Create a durable local file checkpoint.

sb checkpoints

Durable local file checkpoint operations.

sb checkpoints diff

Diff current files against a checkpoint.

Arguments:

  • checkpoint_id (required)

Flags:

  • --json — Output JSON

sb checkpoints list

List durable file checkpoints.

Flags:

  • --limit
  • --json — Output JSON

sb checkpoints show

Show a checkpoint manifest.

Arguments:

  • checkpoint_id (required)

Flags:

  • --json — Output JSON

sb classify

Shortcut for sb tasks classify.

sb claude

Claude Code CLI bridge — run claude inline or as a durable background session.

sb claude check

Check whether the claude binary is available and show its version.

sb claude review

Ask Claude Code to review a file/directory with SecondBrain domain knowledge.

Arguments:

  • path (required)

Flags:

  • --context, -c — Topic for vault context lookup
  • --model, -m — Model override
  • --cwd — Working directory for resolving relative review paths
  • --add-dir — Additional directories to expose to Claude (repeatable)
  • --permission-mode — Claude permission mode override passed to claude --permission-mode
  • --no-context — Skip vault context injection
  • --json — Emit JSON result

sb claude run

Run Claude Code CLI with SecondBrain vault context injected.

Arguments:

  • task (required)

Flags:

  • --context, -c — Topic to fetch from vault for context (default: use task text)
  • --model, -m — Model override passed to claude --model
  • --cwd — Working directory for the Claude session
  • --add-dir — Additional directories to expose to Claude (repeatable)
  • --permission-mode — Claude permission mode override passed to claude --permission-mode
  • --top-k — Number of vault chunks to inject
  • --no-context — Skip vault context injection
  • --session-id — Durable background session id (primarily for --detach runs)
  • --detach — Create a durable background Claude session instead of running inline.
  • --session, -s — Durable bridge session name. Reuses the prior session's workdir and resumes Claude's conversation across turns.
  • --skills — Comma-separated skill slugs to inject under .claude/skills/ for this run.
  • --json — Emit JSON result

sb clip

Capture clipboard contents into SecondBrain.

sb code

Code analysis and generation operations.

sb code analyze

Analyze code structure: functions, classes, imports, complexity.

Arguments:

  • path (required)

Flags:

  • --json — Output as JSON

sb code generate

Generate code using the coder agent (supervised).

Arguments:

  • prompt (required)

Flags:

  • --output, -o — Output file path
  • --agent-profile — Agent profile to use

sb code refactor

Propose refactoring for a file (supervised).

Arguments:

  • path (required)
  • instruction (required)

Flags:

  • --agent-profile — Agent profile to use

sb codebase

Codebase sensor — index, search, and promote local source files into the brain.

sb codebase context

Generate an ambient context document from the codebase index.

Flags:

  • --focus — Filter modules whose dotted name or path contains this substring
  • --out — Write the document to this path (otherwise stdout)
  • --max-packages
  • --samples-per-package
  • --max-top-modules
  • --recent-commits
  • --json — JSON output (forces stdout)

sb codebase docs-index

Build the progressive-disclosure docs index for the docs/ tree.

Flags:

  • --docs-root — docs/ directory (default: /docs)
  • --write — Write INDEX.md to docs_root (otherwise stdout)
  • --exclude — Subdir name to skip (repeatable; default: archive)
  • --json — JSON output (forces stdout)

sb codebase index

Walk the repo, parse Python files, populate the codebase index.

Flags:

  • --root — Repo root to scan (default: configured cwd_root)
  • --limit — Stop after N files (0 = unlimited)
  • --clean — Remove files from index that no longer exist on disk
  • --json — JSON output

sb codebase lsp

LSP bridge — talk to language servers.

sb codebase lsp servers

List installed LSP servers detected on PATH.

Flags:

  • --language — Filter by language
  • --json — JSON output

sb codebase lsp symbols

Fetch symbols for a file via the auto-detected LSP server.

Arguments:

  • file (required)

Flags:

  • --language
  • --json — JSON output

sb codebase promote

Promote module summaries/docstrings into Chitta as code memories.

Flags:

  • --limit — Stop after N files (0 = unlimited)
  • --json — JSON output

Search the symbol index by substring.

Arguments:

  • query (required)

Flags:

  • --kind — Filter by kind (class, function, method, ...)
  • --limit — Max results
  • --json — JSON output

sb codebase status

Show codebase index statistics.

Flags:

  • --json — JSON output

sb codebase working-set

Build a WorkingSet from --path/--symbol flags and print its enriched view.

Flags:

  • --path — Repo-relative file path to mark as touched (repeatable).
  • --symbol — Qualified symbol name to mark as touched (repeatable).
  • --reason — Reason for the touch
  • --markdown — Render compact markdown for LLM injection
  • --json — JSON output

sb codex

Codex CLI bridge — run codex inline or as a durable background session.

sb codex adversarial-review

Run an adversarial design review with Codex.

Arguments:

  • target (required)

Flags:

  • --context, -c — Vault topic to inject as context
  • --cwd — Working directory for the Codex session
  • --add-dir — Extra directories to expose to Codex (repeatable)
  • --model, -m
  • --no-context — Skip vault context injection
  • --top-k
  • --json — Emit JSON result

sb codex bg

Spawn a durable background Codex run and return its session id.

Arguments:

  • task (required)

Flags:

  • --context, -c
  • --approval-mode, -a
  • --model, -m
  • --cwd
  • --top-k
  • --no-context
  • --session-id — Pin a specific background session id
  • --json

sb codex check

Check whether the codex binary is available and show its version.

sb codex rescue

Hand off a failing fix to Codex (default sandbox: workspace-write).

Arguments:

  • path (required)

Flags:

  • --bug, -b — Description of what is broken (be specific).
  • --repro, -r — Optional reproduction command Codex should run first.
  • --approval-mode, -a — suggest | auto-edit | full-auto (default: auto-edit so Codex can write)
  • --context, -c — Vault topic to inject
  • --cwd
  • --add-dir
  • --model, -m
  • --no-context
  • --top-k
  • --json

sb codex result

Print the final output for a completed background Codex session.

Arguments:

  • session_id (required)

Flags:

  • --json

sb codex review

Ask Codex to review a file/directory with SecondBrain domain knowledge.

Arguments:

  • path (required)

Flags:

  • --context, -c — Topic for vault context lookup
  • --approval-mode, -a — suggest | auto-edit | full-auto
  • --cwd — Working directory for resolving relative review paths
  • --add-dir — Additional directories to expose to Codex (repeatable)
  • --no-context — Skip vault context injection
  • --json — Emit JSON result

sb codex run

Run Codex CLI with SecondBrain vault context injected.

Arguments:

  • task (required)

Flags:

  • --context, -c — Topic to fetch from vault for context (default: use task text)
  • --approval-mode, -a — suggest | auto-edit | full-auto
  • --model, -m — Model override passed to codex --model
  • --cwd — Working directory for the Codex session
  • --add-dir — Additional directories to expose to Codex (repeatable)
  • --top-k — Number of vault chunks to inject
  • --no-context — Skip vault context injection
  • --session-id — Durable background session id (primarily for --detach runs)
  • --detach — Create a durable background Codex session instead of running inline.
  • --session, -s — Durable bridge session name. Reuses the workdir and pins CODEX_HOME so files/skills persist across turns.
  • --skills — Comma-separated skill slugs to inject under .codex/skills/ for this run.
  • --json — Emit JSON result

sb codex status

List recent background Codex sessions, or show details for one.

Arguments:

  • session_id

Flags:

  • --limit, -n
  • --json

sb commitments

Durable commitments and follow-up tracking.

sb commitments done

Mark a commitment done.

Arguments:

  • commitment_id (required)

sb commitments list

List commitments, newest first.

Flags:

  • --status, -s — Filter by status
  • --project-id — Filter by linked project id
  • --limit — Maximum commitments to show
  • --json — Output JSON

sb commitments remind

Queue a governed reminder action for one commitment.

Arguments:

  • commitment_id (required)

Flags:

  • --json — Output JSON

sb commitments remind-conversations

Queue governed reminder actions for promoted email promises nearing due date.

Flags:

  • --days-ahead — Queue reminders for conversation-origin commitments due within this many days.
  • --limit — Maximum commitments to consider
  • --dry-run — Preview eligible commitments without queueing reminder actions.
  • --json — Output JSON

sb commitments remind-meetings

Queue governed reminder actions for due-soon meeting commitments.

Flags:

  • --days-ahead — Queue reminders for meeting-origin commitments due within this many days.
  • --limit — Maximum commitments to consider
  • --dry-run — Preview eligible commitments without queueing reminder actions.
  • --json — Output JSON

sb commitments show

Show one commitment.

Arguments:

  • commitment_id (required)

Flags:

  • --json — Output JSON

sb commitments update

Update commitment status.

Arguments:

  • commitment_id (required)

Flags:

  • --status — New status: open|in_progress|done|dropped

sb company-brain

Company Brain - compile operating knowledge into executable agent skill drafts.

sb company-brain answer

Workspace-scoped retrieval-grounded answer.

Arguments:

  • workspace_id (required)
  • question (required)

Flags:

  • --top-k
  • --min-authority — Minimum SourceAuthority floor (1=scratch, 8=policy).
  • --json

sb company-brain baselines

List named Company Brain baselines stored under the local state directory.

Flags:

  • --json

sb company-brain compile

Compile company operating knowledge into cited agent skill drafts.

Arguments:

  • target

Flags:

  • --company — Company/workspace name for the compiled report.
  • --glob — Markdown glob for directory inputs.
  • --recursive, --flat — Recurse through directory inputs.
  • --max-files — Maximum Markdown files to inspect.
  • --max-entries — Maximum operating-map entries to emit.
  • --format — Output format.
  • --json — Shortcut for --format json.
  • --out — Write output to this file.
  • --mirror-to-workspace — Also mirror the compiled report into the named Company Brain workspace's substrate (MemoryStore + DecisionCatalog). Idempotent.

sb company-brain context-build

Workspace-scoped context packs.

sb company-brain context-build run

Build a ContextPackV2 scoped to the workspace.

Arguments:

  • workspace_id (required)

Flags:

  • --task — Task / intent driving the pack.
  • --top-k
  • --json

sb company-brain diff

Diff a saved Company Brain report against current Markdown sources.

Arguments:

  • baseline (required)
  • target

Flags:

  • --company
  • --glob
  • --recursive, --flat
  • --max-files
  • --max-entries
  • --format
  • --json
  • --out
  • --fail-on-drift, --no-fail-on-drift — Exit non-zero when semantic Company Brain drift is detected.

sb company-brain ingest

Workspace ingestion via the per-workspace memory store.

sb company-brain ingest from-dir

Recursively ingest a directory through brain/ingest/parser + chunker.

Arguments:

  • workspace_id (required)
  • root (required)

Flags:

  • --glob
  • --kind
  • --max-files
  • --chunk-size
  • --chunk-overlap
  • --json

sb company-brain ingest run

Ingest one source unit into the workspace's memory + evidence stores.

Arguments:

  • workspace_id (required)
  • source_ref (required)

Flags:

  • --kind
  • --from-file
  • --content
  • --anchor
  • --no-memory
  • --json

sb company-brain init

Register a workspace and materialise its dedicated SQLite stores.

Arguments:

  • workspace_id (required)

Flags:

  • --name — Human-readable name.
  • --root — Source root directory.
  • --connectors — Comma-separated connector list.
  • --sensitivity-floor — 0=public, 1=internal, 2=confidential, 3=restricted.
  • --owner
  • --description
  • --json

sb company-brain skills

Company Brain skill operations.

sb company-brain skills validate

Validate a kernel AgentDefinition YAML used as a Company Brain skill.

Arguments:

  • agent_yaml (required)

Flags:

  • --strict, --lenient
  • --json

sb company-brain skills-write

Write compiled Company Brain skill drafts as SKILL.md folders.

Arguments:

  • target

Flags:

  • --out — Directory where generated SKILL.md folders should be written.
  • --company
  • --glob
  • --recursive, --flat
  • --force — Overwrite existing skills.
  • --json

sb company-brain snapshot

Persist the current Company Brain compile as a named local baseline.

Arguments:

  • target

Flags:

  • --name — Named baseline.
  • --company
  • --glob
  • --recursive, --flat
  • --max-files
  • --max-entries
  • --json

sb company-brain workspaces

Manage Company Brain workspaces.

sb company-brain workspaces list

List registered Company Brain workspaces.

Flags:

  • --json

sb company-brain workspaces remove

Remove a workspace from the registry (storage stays unless --delete-storage).

Arguments:

  • workspace_id (required)

Flags:

  • --delete-storage — Also remove the workspace's storage directory on disk.

sb concierge

Travel Concierge — memory-aware planner with adaptive replanning.

sb concierge pack

Print the Trip Pack for an existing concierge session.

Flags:

  • --session, -s — Session ID to retrieve trip pack for
  • --json — Output raw JSON
  • --output, -o — Write to file

sb concierge plan

Plan a trip with preference memory, policy validation, and a Trip Pack.

Arguments:

  • query (required)

Flags:

  • --user, -u — User ID for memory recall
  • --session, -s — Existing session ID to continue (optional)
  • --diagnostics — Include cognitive trace
  • --json — Output raw JSON
  • --output, -o — Write trip pack to file

sb concierge replan

Replan a trip after a disruption — the 'agentic trip fixer'.

Arguments:

  • disruption (required)

Flags:

  • --session, -s — Session ID from a previous sb concierge plan run
  • --json — Output raw JSON
  • --output, -o — Write updated trip pack to file

sb concierge sessions

List all concierge sessions.

Flags:

  • --user, -u — Filter by user ID
  • --json — Output raw JSON

sb config

Configuration management operations.

sb config edit

Open the configuration file in $EDITOR.

sb config init

Generate a starter config.yaml with documented defaults.

Flags:

  • --force, -f — Overwrite existing config

sb config path

Print the configuration file path.

sb config set

Update a single configuration value in the YAML file.

Arguments:

  • key (required)
  • value (required)

sb config show

Display all effective configuration values with their source.

sb connectors

Connector, backfill, and replay operations.

sb connectors auth-delete

Arguments:

  • name (required)

sb connectors auth-list

Flags:

  • --json

sb connectors auth-set

Arguments:

  • name (required)

Flags:

  • --base-url
  • --provider
  • --token-env
  • --auth-scheme
  • --header — Repeatable KEY=VALUE
  • --query-param — Repeatable KEY=VALUE

sb connectors auth-show

Arguments:

  • name (required)

Flags:

  • --json

sb connectors auth-test

Arguments:

  • name (required)
  • path (required)

Flags:

  • --method
  • --json

sb connectors export-events

Arguments:

  • output (required)

Flags:

  • --event-type
  • --limit

sb connectors github-sync

Sync GitHub issues/wiki into the vault (incremental).

Flags:

  • --json
  • --write-vault, --no-write-vault
  • --what — Comma-separated: issues,wiki,discussions

sb connectors m365-device-login

Sign in with Microsoft identity device-code flow for Microsoft 365 sync.

Flags:

  • --client-id — Microsoft Entra app client ID.
  • --tenant — Tenant, domain, or organizations.
  • --scope — Graph delegated scope. Repeat or use config defaults.
  • --token-env — Keychain token name to store.
  • --json

sb connectors m365-sync

Sync Microsoft 365 mail, calendar, OneDrive files, and Teams into the vault.

Flags:

  • --source, -s — Source to sync: mail, calendar, files, teams. Repeat or comma-separate.
  • --mail-query
  • --mail-folder
  • --limit — Primary per-source limit for mail, calendar, and files.
  • --chats-limit
  • --chat-messages
  • --team-channel — Teams channel as TEAM_ID/CHANNEL_ID or TEAM_ID|CHANNEL_ID. Repeatable.
  • --channel-messages
  • --days-back
  • --days-ahead
  • --download-files, --metadata-only — Download supported OneDrive files instead of writing metadata notes only.
  • --max-file-bytes
  • --max-pages
  • --dry-run
  • --write-vault, --no-write-vault
  • --ingest, --no-ingest
  • --force
  • --json

sb connectors notion-sync

Sync Notion pages into the vault (incremental).

Flags:

  • --json
  • --write-vault, --no-write-vault

sb connectors replay-events

Arguments:

  • input_path (required)

sb connectors status

Health check for all configured connectors.

Flags:

  • --json

sb context

Codified-context registry and drift operations.

sb context annotate

Write or update a dataset/tool annotation card in ops/cards/.

Arguments:

  • topic (required)

Flags:

  • --title, -t — Human-readable title
  • --description, -d — What this dataset/tool represents
  • --grain, -g — Row grain / unit of analysis (e.g. 'one row per event')
  • --caveats — Known gotchas, filters, edge cases
  • --source, -s — Source URI (table name, file path, API, …)
  • --keywords, -k — Comma-separated keywords for retrieval
  • --layer — Context layer (ANNOTATION, INSTITUTIONAL, …)
  • --edit — Open in $EDITOR after writing
  • --json — Output JSON

sb context cards

List all annotation cards in ops/cards/.

Flags:

  • --json — Output JSON

sb context compile

Compile runtime context from promoted knowledge first and raw evidence second.

Arguments:

  • query (required)

Flags:

  • --json — Output JSON

sb context doctor

Inspect context freshness, vector coverage, promoted artifacts, and memory indexes.

Flags:

  • --write-rank-hints — Persist retrieval rank hints from pruning diagnostics
  • --cleanup-rank-hints — Preview cleanup of resolved context-doctor rank hints
  • --apply — Apply --cleanup-rank-hints; default is dry run
  • --json — Output JSON

sb context drift

Detect context drift (source changed but subsystem specs unchanged).

Flags:

  • --base-ref — Git base ref (default from config)
  • --strict — Exit non-zero when stale subsystem drift is detected
  • --json — Output JSON

sb context explain

Explain context selection and budget events recorded for a run.

Arguments:

  • run_id (required)

Flags:

  • --json — Output JSON

sb context files

List files mapped to one codified-context subsystem.

Arguments:

  • subsystem_id (required)

Flags:

  • --json — Output JSON

sb context find

Find relevant codified-context subsystems for a task.

Arguments:

  • task (required)

Flags:

  • --top-k, -k
  • --json — Output JSON

sb context index

Compile the offline context index from promoted knowledge plus bounded evidence.

Flags:

  • --json — Output JSON

sb context sources

List sources that contributed to the offline context index.

Flags:

  • --layer, -l — Optional layer filter (ANNOTATION, CODE_ENRICHMENT, INSTITUTIONAL, DECISION, MEMORY, RUNTIME, USAGE)
  • --limit — Maximum rows to show
  • --items — Show individual context items instead of grouping by source
  • --json — Output JSON

sb context subsystems

List codified-context subsystems from registry.

Flags:

  • --json — Output JSON

sb context suggest-agent

Suggest best profile for a task based on codified context.

Arguments:

  • task (required)

Flags:

  • --json — Output JSON

sb daemon

SB daemon — long-running background worker (sweeper, task queue, harness refresh).

sb daemon fleet

List per-host serve-daemon heartbeats (alive / stale).

Flags:

  • --json — Emit JSON.
  • --stale-after — Seconds after which a heartbeat counts as stale.

sb daemon logs

Tail the daemon log file.

Flags:

  • --follow, -f — Stream new log lines (tail -f).
  • --lines, -n — Show last N lines.

sb daemon restart

Restart the SB daemon, starting it if it is not running.

Flags:

  • --tick-seconds, -t — Supervisor tick interval (seconds) for the new daemon.
  • --wait — Seconds to wait for first heartbeat before returning.
  • --timeout — Seconds to wait for graceful stop before starting again.
  • --force — Send SIGKILL instead of SIGTERM when stopping the current daemon.

sb daemon start

Start the SB daemon (detaches to background by default).

Flags:

  • --foreground, -f — Run in the current terminal instead of detaching.
  • --tick-seconds, -t — Supervisor tick interval (seconds).
  • --wait — Seconds to wait for first heartbeat before returning.

sb daemon status

Show daemon runtime status + worker stats.

Flags:

  • --json — Emit JSON.

sb daemon stop

Stop the running daemon (SIGTERM; --force for SIGKILL).

Flags:

  • --timeout, -t — Seconds to wait for graceful SIGTERM.
  • --force — Send SIGKILL instead of SIGTERM.

sb daemon tick

Run one tick of the default worker set synchronously (for dev/test).

Flags:

  • --json — Emit JSON.

sb daily

Daily log — view and append to vault/daily/YYYY-MM-DD.md.

sb daily append

Append a note to today's daily log.

Arguments:

  • text (required)

Flags:

  • --section, -s — Section heading

sb daily show

Show today's (or a specified date's) daily log.

Flags:

  • --date, -d — Date to show (YYYY-MM-DD). Defaults to today.

sb data-agent

Data Agent — advanced grounded answers with explicit plan/debug output.

sb data-agent ask

Advanced question-answering interface with full context, planning, and debug detail.

Arguments:

  • question (required)

Flags:

  • --provider, -p — LLM provider (anthropic, ollama, …)
  • --model, -m — Model name override
  • --top-k, -k — Maximum runtime context items to retrieve
  • --budget — Token budget for context
  • --json — Emit JSON result
  • --debug — Show debug info
  • --no-sql — Skip SQL execution even if detected
  • --agentic, --no-agentic — Enable agentic planning/execution loop (default: on)
  • --max-iterations — Maximum planner/repair iterations for agentic mode
  • --llm-intent, --heuristic-intent — Use LLM-assisted hybrid intent classification

sb data-agent plan

Show the execution plan without running it.

Arguments:

  • question (required)

sb data-agent save-learning

Save a correction or learning to the memory system.

Arguments:

  • content (required)

Flags:

  • --scope, -s — Memory scope: global (shared) or personal (session-bound)
  • --category, -c — Memory category tag (e.g. data_agent.correction, query.pattern)

sb data-agent status

Show advanced answer-path readiness: promoted knowledge, index, provider, and memory.

sb dataset

Versioned dataset management.

sb dataset clean

Clean a dataset — dedup, null-fill, and whitespace normalisation.

Arguments:

  • file (required)

Flags:

  • --output, -o — Output CSV path.
  • --no-dedup — Skip deduplication.
  • --fill-null — Replace null/empty cells with this string.
  • --no-strip — Skip whitespace stripping.
  • --json — Output JSON summary.

sb dataset join-check

Analyse join coverage between two datasets before merging.

Arguments:

  • left (required)
  • right (required)

Flags:

  • --on — Column name to join on.
  • --json — Output raw JSON report.
  • --report — Generate a structured report.
  • --format, -f — Report format: md, html, docx.
  • --output, -o — Write output to path.

sb dataset list

List all versioned datasets.

sb dataset profile

Profile a dataset file — null rates, types, candidate keys, and data-quality caveats.

Arguments:

  • file (required)

Flags:

  • --json — Output raw JSON profile.
  • --report — Generate a structured report instead of a table.
  • --format, -f — Report format: md, html, docx (only with --report).
  • --output, -o — Write report/JSON to this path.

sb dataset seed

Seed the qa-baseline dataset from vault content.

Flags:

  • --force — Seed even if qa-baseline already exists

sb dataset show

Show examples from a dataset version.

Arguments:

  • name (required)

Flags:

  • --version, -v — Version (default: latest)
  • --limit, -n — Max examples to show

sb decide

Structured decision memo with memory, tradeoffs, and guardrails.

sb decision-extract

Extract decisions and action items from a markdown file.

sb decisions

Decision catalog operations.

sb decisions emit

Emit a runtime decision record.

Flags:

  • --key — Decision key (e.g. mkt.audience.eligibility)
  • --version — Decision version
  • --subject-ids — JSON list of subject IDs
  • --output — JSON dict with 'selected' key
  • --confidence — Confidence score 0.0–1.0
  • --reason-codes — Comma-separated reason codes
  • --status — DECIDED|DEFERRED|REJECTED|ESCALATED
  • --run-id — Lineage run ID
  • --intent-ref — Link to goal or intent that triggered this decision
  • --strategy — Logic path chosen (e.g. direct_answer, synthetic_reasoning)
  • --tags — Comma-separated category labels (e.g. architecture,security)
  • --regret-check — Reference to past failures that influenced this decision
  • --vault-mirror, --no-vault-mirror — Also write a markdown mirror to vault

sb decisions extract

Extract decision records into the local decision catalog.

Arguments:

  • target (required)

Flags:

  • --project — Optional project label
  • --glob — Glob used when target is a directory
  • --recursive — Recurse into subdirectories
  • --json — Output batch report as JSON

sb decisions keys

List all known runtime decision keys.

sb decisions list

List catalogued decisions or runtime decision records.

Flags:

  • --project — Project filter for catalog decisions
  • --key — Filter by runtime decision key (e.g. mkt.guardrail.evaluation)
  • --limit — Max rows
  • --since — ISO datetime lower bound (for runtime decisions)

sb decisions meeting

Extract decisions/actions/questions from a meeting transcript.

Arguments:

  • transcript_file (required)

Flags:

  • --project — Optional project label
  • --project-id — Optional project id for scheduled follow-up tasks
  • --meeting-title — Meeting title shown in generated artifacts
  • --meeting-date — Meeting date (YYYY-MM-DD). Defaults to current date.
  • --engine — Workflow engine: auto|local|adk
  • --adk-model — ADK model when --engine adk (or auto with ADK available).
  • --ping-owners, --no-ping-owners — Generate owner ping draft file grouped by assignee
  • --a2a-pings, --no-a2a-pings — Also publish owner pings to local A2A queue

sb decisions namespaces

Show the canonical decision-key namespace registry.

sb decisions policy-list

Flags:

  • --status

sb decisions policy-upsert

Arguments:

  • policy_id (required)

Flags:

  • --strategy
  • --min-confidence
  • --require-evidence, --no-require-evidence
  • --require-human-approval, --no-require-human-approval

sb decisions retrospective

Review downstream execution signals for one persisted decision.

Arguments:

  • selector

Flags:

  • --project — Optional project filter
  • --limit — Max decision rows to inspect
  • --json — Output JSON

sb decisions show

Show full details of a runtime decision record.

Arguments:

  • record_id (required)

sb decisions stats

Show statistics for a decision key.

Arguments:

  • decision_key (required)

sb dev

Developer workflow operations.

sb dev review

Run formatting, lint, type checks, and tests with prioritized findings.

Flags:

  • --timeout-s — Per-step timeout seconds
  • --json — Output JSON

sb dev silent-failures

Surface swallowed-exception / no-log handlers (see brain/dev/silent_failures.py).

Flags:

  • --scan — Source root to scan (repeatable). Default: brain/ under the repo root.
  • --root — Repo root used for relative-path reporting.
  • --min-severity — Only report findings at or above this severity (low|medium|high).
  • --limit — Max findings to render in the table (use --json for full output).
  • --check-baseline, --no-check-baseline — After scanning, compare counts against tools/silent_failures_baseline.json.
  • --json — Output JSON.

sb dev test-quality

Surface empty / no-assertion / unexplained-skip tests (see brain/dev/test_quality.py).

Flags:

  • --scan — Test root to scan (repeatable). Default: tests/ under the repo root.
  • --root — Repo root used for relative-path reporting.
  • --min-severity — Only report findings at or above this severity (low|medium|high).
  • --limit — Max findings to render in the table (use --json for full output).
  • --check-baseline, --no-check-baseline — After scanning, compare counts against tools/test_quality_baseline.json.
  • --json — Output JSON.

sb diff

Show git diff, or sb diff <checkpoint_id> for file checkpoints.

sb dify

Dify AI apps — invoke external AI flows as governed tools.

sb dify chat

Send a message to a Dify app through ToolExecutor and ToolPolicy.

Arguments:

  • message (required)

Flags:

  • --app-id — Dify app ID ('default' uses SB_DIFY_; named apps use SB_DIFY_APP__)
  • --inputs — Extra inputs as JSON
  • --conv-id — Resume conversation
  • --json — Output JSON

sb dify registered

Show Dify apps registered as SecondBrain tools.

Flags:

  • --json — Output JSON

sb dify status

Check Dify daemon connectivity and configuration.

Flags:

  • --json — Output JSON status

sb dify workflow

Invoke a Dify workflow app through ToolExecutor and ToolPolicy.

Arguments:

  • app_id (required)

Flags:

  • --input, -i — JSON inputs
  • --json — Output JSON

sb digest

Proactive briefings from the second brain.

sb digest morning

Today's briefing: stale commitments, pending actions, capture pulse.

Flags:

  • --vault, -v — Vault path (defaults to settings.paths.vault).
  • --stale-days — Days before a commitment counts as stale.
  • --top — Max stale commitments to surface.
  • --capture-window — Days of capture activity to summarise.

sb discord

Discord — send messages via webhook or bot.

sb discord auth

Save Discord webhook URL and/or bot token to keys.env.

sb discord listen

Start a Discord bot and listen for messages (blocking).

sb discord send

Send a message to Discord via webhook or bot.

Arguments:

  • message (required)

Flags:

  • --channel, -c — Channel ID (bot mode only)
  • --username — Bot display name for webhook
  • --force — Send even if proactivity < assistant

sb discord test

Test the Discord webhook connection.

sb docs

Auto-generated reference docs for CLI, providers, and decisions.

sb docs all

Run every reference-doc generator (cli + providers + decisions)

Flags:

  • --write, --no-write — Write generated docs to disk (default: yes).

sb docs check

Diff each generator's output against the committed file. Exits 1 on drift.

sb docs cli-reference

Render the full sb CLI reference from the live cli_schema.json.

Flags:

  • --write — Write to docs/reference/cli.md (else stdout).

sb docs decisions

Render the decision-namespace registry as a Markdown reference.

Flags:

  • --write — Write to docs/reference/decisions.md (else stdout).

sb docs providers

Render the provider registry as a Markdown reference.

Flags:

  • --write — Write to docs/reference/providers.md (else stdout).

sb doctor

Environment health and repair operations.

sb doctor split-state

Inspect or apply the split state-database topology.

Flags:

  • --apply — Create the split DBs, migrate legacy events.db data, and archive it.
  • --check — Show split-state status without applying migration.

sb drafts

Draft management — generate, list, and send message drafts.

sb drafts approvals

List pending draft-send approvals.

Flags:

  • --json — Output JSON

sb drafts approve-send

Approve and execute a queued draft-send action.

Arguments:

  • action_id (required)

Flags:

  • --approver — Approver label
  • --why — Optional rationale
  • --json — Output JSON

sb drafts decision-memo

Create a decision memo draft from the persisted decision catalog.

Arguments:

  • selector

Flags:

  • --to — Recipient email or label
  • --project — Optional project filter for decision lookup
  • --limit — Maximum candidate decisions to inspect
  • --no-llm — Skip LLM polishing and use deterministic body
  • --json — Output JSON

sb drafts deny-send

Deny a queued draft-send action.

Arguments:

  • action_id (required)

Flags:

  • --approver — Approver label
  • --why — Optional rationale
  • --json — Output JSON

sb drafts expire

Expire stale drafts that have passed their expiry date.

sb drafts generate

Generate a draft email using LLM in your voice.

Flags:

  • --to — Recipient email or name
  • --subject — Email subject
  • --context, -c — Context topic for the draft
  • --no-llm — Skip LLM; create empty draft

sb drafts history

Show governed send history for one draft.

Arguments:

  • draft_id (required)

Flags:

  • --json — Output JSON

sb drafts inbox-reply

Create an inbox reply draft from Gmail message and thread context.

Arguments:

  • selector

Flags:

  • --to — Override recipient; defaults to the selected sender
  • --query — Gmail query used to find candidate messages
  • --max-messages — Maximum candidate messages to inspect
  • --no-llm — Skip LLM polishing and use deterministic body
  • --json — Output JSON

sb drafts list

List drafts by status.

Flags:

  • --status, -s — active | sent | expired | all

sb drafts meeting-followup

Create a meeting follow-up draft from the latest persisted closure.

Arguments:

  • selector

Flags:

  • --to — Recipient email or label
  • --project-id — Optional project id to narrow the meeting search
  • --subject-prefix — Subject prefix for the draft
  • --no-llm — Skip LLM polishing and use deterministic body
  • --json — Output JSON

sb drafts project-update

Create a project update draft from persisted project continuity.

Arguments:

  • project_id (required)

Flags:

  • --to — Recipient email or label
  • --decision-limit — Maximum recent decisions to include
  • --meeting-limit — Maximum meeting records to inspect
  • --task-limit — Maximum linked commitments/tasks to inspect
  • --no-llm — Skip LLM polishing and use deterministic body
  • --json — Output JSON

sb drafts requeue-send

Queue a fresh governed send after the latest one was denied.

Arguments:

  • draft_id (required)

Flags:

  • --json — Output JSON

sb drafts send

Send a draft via Gmail and move it to sent/.

Arguments:

  • draft_id (required)

Flags:

  • --force — Send even if proactivity level < assistant
  • --governed — Queue a governed send action for approval instead of sending immediately
  • --dry-run — Print what would be sent, don't send
  • --json — Output JSON

sb drafts show

Show a draft's full content.

Arguments:

  • draft_id (required)

sb embeddings

Embedding providers — list, probe, benchmark across Voyage/OpenAI/Cohere/Jina/local.

sb embeddings benchmark

Compare providers on retrieval or memory benchmark goldens.

Flags:

  • --goldens — JSONL file with retrieval goldens
  • --memory-fixture — Memory benchmark JSON/JSONL/directory. Converts LoCoMo, LongMemEval, BEAM, or synthetic cases into embedding retrieval goldens.
  • --memory-suite — Memory fixture suite: auto|synthetic|locomo|longmemeval|beam
  • --max-cases — Limit memory-fixture cases for large public benchmark tiers.
  • --providers — Comma-separated list of providers to compare, or 'auto' for all available
  • --top-k
  • --json

sb embeddings list

List all known embedding providers and the active selection.

sb embeddings migrate

Re-encode all known collections under the target provider.

Flags:

  • --to — Target provider (voyage|openai|cohere|jina|bge-m3|st|ollama|hash)
  • --yes — Skip confirmation prompt

sb embeddings probe

Encode a sample with the active provider and print metadata.

Arguments:

  • text

sb env

RL-style local environment episodes and fixtures.

sb env episodes

List persisted environment episodes.

Flags:

  • --env-id — Filter by environment id
  • --status — Filter by episode status
  • --limit — Maximum rows to return
  • --json — Output JSON

sb env export

Export one persisted environment episode for replay or trainer ingestion.

Arguments:

  • episode_id (required)

Flags:

  • --format — Export format: episode, steps-jsonl, or openenv-json
  • --output, -o — Write export to a file

sb env list

List built-in local environment fixtures.

Flags:

  • --json — Output JSON

sb env replay

Replay a persisted episode into a fresh local environment.

Arguments:

  • episode_id (required)

Flags:

  • --workspace-root — Optional local root for replaying workspace env actions
  • --json — Output JSON

sb env run

Run a deterministic fixture episode and optionally persist it.

Arguments:

  • env_id

Flags:

  • --start — Initial counter value
  • --target — Target counter value
  • --amount — Increment amount per step
  • --max-steps — Step budget
  • --target-path — Workspace target file path
  • --answer — Workspace answer text to write and submit
  • --workspace-root — Optional local root for workspace env actions
  • --task-file — JSON/YAML environment task manifest. When set, manifest env_id wins.
  • --no-store — Do not persist the episode
  • --json — Output JSON

sb env show

Show one persisted environment episode with steps.

Arguments:

  • episode_id (required)

Flags:

  • --json — Output JSON

sb eval

Evaluation operations.

sb eval artifact-report

Show the latest artifact evaluation report.

Flags:

  • --suite-prefix
  • --json — Output JSON

sb eval artifact-run

Run deterministic artifact evaluation and persist outcomes in SQLite.

Arguments:

  • paths (required)

Flags:

  • --suite — Suite name
  • --json — Output JSON

sb eval patterns

Pattern golden eval suite.

sb eval patterns list

List all available golden tasks.

Flags:

  • --pattern, -p — Filter by pattern type
  • --json — Emit JSON

sb eval patterns run

Run all (or selected) pattern golden tasks.

Flags:

  • --provider, -p — LLM provider name
  • --id — Run specific golden IDs (repeatable)
  • --json — Emit JSON output
  • --fail-fast — Stop on first failure

sb eval run

Run result-based golden evals and persist outcomes in SQLite.

Flags:

  • --goldens-dir — Golden eval directory
  • --suite — Suite name for persisted metadata
  • --json — Output JSON

sb experiment

Experiment tracking and comparison.

sb experiment compare

Compare two experiment runs side-by-side.

Arguments:

  • id_a (required)
  • id_b (required)

sb experiment list

List saved experiment runs.

Flags:

  • --limit, -n

sb experiment run

Run an experiment: evaluate a task against a dataset with built-in evaluators.

Flags:

  • --dataset, -d — Dataset name
  • --version, -v — Dataset version (default: latest)
  • --task, -t — Task to evaluate: 'ask'
  • --name, -n — Label for this run (auto-generated if omitted)
  • --top-k — Retrieval top-k

sb experiment show

Show detailed results for an experiment run.

Arguments:

  • experiment_id (required)

Flags:

  • --limit, -n — Max results to show

sb explain

Shortcut for sb tasks explain.

sb explain-permissions

Explain effective permission modes, trust boundaries, and approval gates.

sb export-pack

Export the compatibility Context Pack view in JSON or Markdown format.

sb friction

Friction detector — scan for accumulating issues (stale tasks, old open loops, gaps).

sb friction probes

List available friction probes.

sb friction scan

Scan for accumulating friction (stale tasks, old loops, gaps).

Flags:

  • --json — Emit JSON.
  • --min — Minimum severity (info | warning | high).
  • --only — Comma-separated probe names to run.

sb gateway

Gateway runtime operations.

sb gateway approvals

Approval lifecycle operations.

sb gateway approvals expire

Expire stale pending approval requests.

Flags:

  • --max-age-hours — Expire pending requests older than this.
  • --json — Emit machine-readable JSON output.

sb gateway approvals requests

List approval requests.

Flags:

  • --status — Filter by status (pending/approved/denied/expired).
  • --limit — Maximum rows to return.
  • --json — Emit machine-readable JSON output.

sb gateway approvals revoke

Revoke an approval rule.

Arguments:

  • rule_id (required)

sb gateway approvals rules

List active approval rules.

Flags:

  • --scope — Optional scope filter (agent/person/global).
  • --scope-id — Optional scope id filter.
  • --limit — Maximum rows to return.
  • --json — Emit machine-readable JSON output.

sb gateway config-summary

Print effective gateway auth, approval, and unsafe settings.

Flags:

  • --config — Path to gateway YAML/JSON config (default: ~/.secondbrain/gateway.yaml).
  • --json — Emit machine-readable JSON output.

sb gateway doctor

Run gateway preflight diagnostics.

Flags:

  • --config — Path to gateway YAML/JSON config (default: ~/.secondbrain/gateway.yaml).
  • --json — Emit machine-readable JSON output.

sb gateway events

Session event inspection operations.

sb gateway events approvals

List recent approval lifecycle events.

Flags:

  • --limit — Maximum rows to return.
  • --type — Filter by event type: all, approval_request, approval_grant, approval_deny.
  • --session-id — Optional session id filter.
  • --request-id — Optional approval request id filter.
  • --tool-name — Optional tool name filter.
  • --since-hours — Optional lookback window in hours (0 disables).
  • --json — Emit machine-readable JSON output.

sb gateway events diagnostics

List recent runtime diagnostics events.

Flags:

  • --limit — Maximum rows to return.
  • --session-id — Optional session id filter.
  • --provider — Optional provider filter.
  • --model — Optional model filter.
  • --since-hours — Optional lookback window in hours (0 disables).
  • --compaction-only — Show only diagnostics where compaction was triggered.
  • --json — Emit machine-readable JSON output.

sb gateway events errors

List recent runtime/heartbeat error events.

Flags:

  • --limit — Maximum rows to return.
  • --type — Filter by event type: all, runtime_error, heartbeat_error.
  • --session-id — Optional session id filter.
  • --since-hours — Optional lookback window in hours (0 disables).
  • --json — Emit machine-readable JSON output.

sb gateway events export

Export incident-related events as JSONL.

Flags:

  • --limit — Maximum rows to export.
  • --profile — Optional type profile: incident, runtime, approvals, compaction.
  • --types — Comma-separated event types to export, or 'all'. Allowed: runtime_error, heartbeat_error, approval_request, approval_grant, approval_deny, runtime_diagnostics, system_summary.
  • --session-id — Optional session id filter.
  • --request-id — Optional payload request_id filter (applies after SQL fetch).
  • --since-hours — Lookback window in hours (0 disables).
  • --from-event-id — Optional inclusive lower bound for event_id (0 disables).
  • --to-event-id — Optional inclusive upper bound for event_id (0 disables).
  • --include-meta — Prepend a bundle metadata JSONL record.
  • --output — Output path for JSONL ('-' for stdout).
  • --gzip — Compress file output as gzip (.gz).
  • --redact, --raw — Redact sensitive payload values in exported events.
  • --json — Emit machine-readable export summary.

sb gateway events summaries

List recent system summary events.

Flags:

  • --limit — Maximum rows to return.
  • --session-id — Optional session id filter.
  • --since-hours — Optional lookback window in hours (0 disables).
  • --json — Emit machine-readable JSON output.

sb gateway identity

Identity linking operations.

Create or update an identity mapping.

Arguments:

  • person_id (required)

Flags:

  • --channel — Channel name.
  • --peer-id — External peer id on the channel.
  • --config — Path to gateway YAML/JSON config (default: ~/.secondbrain/gateway.yaml).
  • --verified, --unverified — Mark mapping as verified.

sb gateway identity list

List identity mappings.

Flags:

  • --config — Path to gateway YAML/JSON config (default: ~/.secondbrain/gateway.yaml).
  • --person-id — Optional person id filter.
  • --channel — Optional channel filter.
  • --limit — Maximum rows to return.
  • --json — Emit machine-readable JSON output.

Remove an identity mapping.

Flags:

  • --channel — Channel name.
  • --peer-id — External peer id on the channel.
  • --config — Path to gateway YAML/JSON config (default: ~/.secondbrain/gateway.yaml).

sb gateway init

Scaffold gateway config, agents, constitution, and heartbeat stubs.

Flags:

  • --config — Path to write gateway config (default: ~/.secondbrain/gateway.yaml).
  • --force — Overwrite existing files.

sb gateway outbound

Outbound queue dead-letter operations.

sb gateway outbound list

List outbound deliveries.

Flags:

  • --status — Filter by status (pending/sending/failed/delivered). Empty string lists all.
  • --limit — Maximum rows to return.
  • --json — Emit machine-readable JSON output.

sb gateway outbound purge

Delete old terminal outbound rows.

Flags:

  • --status — Terminal status to purge: delivered or failed.
  • --older-than-hours — Delete rows older than this age in hours.
  • --json — Emit machine-readable JSON output.

sb gateway outbound retry

Requeue failed outbound deliveries.

Flags:

  • --queue-id — Specific failed queue id to requeue (repeatable).
  • --all-failed — Requeue all failed deliveries.
  • --reset-attempts, --keep-attempts — Reset attempts to 0 when requeued.
  • --immediate, --with-backoff — Retry immediately or with base backoff.
  • --json — Emit machine-readable JSON output.

sb gateway run

Run gateway service with configured channels.

Flags:

  • --config — Path to gateway YAML/JSON config (default: ~/.secondbrain/gateway.yaml).

sb gcal

Google Calendar — read upcoming events via OAuth2.

sb gcal auth

Authenticate with Google Calendar via OAuth2 browser flow.

Flags:

  • --no-browser — Use console auth flow (headless environments)

sb gcal calendars

List all available Google Calendars.

Flags:

  • --json — Output raw JSON

sb gcal events

List upcoming Google Calendar events.

Flags:

  • --days, -d — Number of days ahead to look
  • --calendar, -c — Calendar ID
  • --all — Fetch from all configured calendars
  • --json — Output raw JSON

sb gcal ingest

Ingest upcoming Google Calendar events into SecondBrain memory.

Flags:

  • --days, -d — Days ahead to ingest
  • --all — All configured calendars

sb gmail

Gmail — read and send email via Google OAuth2.

sb gmail auth

Authenticate with Gmail via OAuth2 browser flow.

Flags:

  • --no-browser — Use console auth flow (headless environments)

sb gmail fetch

Fetch full details of a Gmail message.

Arguments:

  • message_id (required)

Flags:

  • --json — Output raw JSON

sb gmail ingest

Ingest recent Gmail messages into SecondBrain memory.

Flags:

  • --query, -q — Gmail search query (e.g. 'is:unread')
  • --max, -n — Max messages to ingest
  • --label — Optional label filter (e.g. 'INBOX')

sb gmail list

List recent Gmail messages.

Flags:

  • --query, -q — Gmail search query
  • --max, -n — Maximum messages to list
  • --json — Output raw JSON

sb gmail profile

Show the authenticated Gmail account profile.

sb governance

Inspect and operate the live agent governance gate.

sb governance charter

Print the versioned charter (allowed / denied intents).

sb governance killswitch

Break-glass kill switch (durable across processes).

sb governance killswitch engage

Stop a scope now. * halts the whole agent.

Arguments:

  • scope

Flags:

  • --reason — Why the switch is engaged

sb governance killswitch list

List engaged kill-switch scopes.

sb governance killswitch release

Disengage the kill switch for a scope.

Arguments:

  • scope

sb governance release

Print the pinned release tuple (prompt / model / policy / tools / memory).

sb governance retention

Purge events / chat messages older than the retention window (control #14).

Flags:

  • --days — Override policies.log_retention_days for this run
  • --apply — Actually delete (default is a dry-run preview)

sb governance status

Show the governing charter, enforcement mode, and engaged kill-switch scopes.

sb graph

Context graph operations.

sb graph build

Build local context graph from markdown vault content.

Flags:

  • --path — Vault path to build graph from

sb graph enrich

Enrich the graph from structured decisions, projects, commitments, and actions.

Flags:

  • --json — Output JSON

sb graph query

Query local context graph.

Arguments:

  • query (required)

Flags:

  • --limit — Maximum rows

sb graph show

Show a graph node with neighbors.

Arguments:

  • node_id (required)

sb grounded

Grounded reasoning — KARL-inspired closed-corpus QA, eval, and synthesis.

sb grounded answer

Answer a question using grounded reasoning over a local corpus.

Arguments:

  • question (required)

Flags:

  • --corpus-id — Corpus to search
  • --top-k — Number of chunks per retrieval
  • --max-steps — Maximum reasoning steps
  • --provider — LLM provider name
  • --model — Model name
  • --parallel, -p — Number of parallel rollouts (0=disabled)
  • --trace — Show full reasoning trace
  • --json — Output JSON

sb grounded eval

Run the grounded reasoning eval suite.

Flags:

  • --tasks-dir — Path to eval tasks directory
  • --provider — LLM provider name
  • --model — Model name
  • --json — Output JSON

sb grounded replay

Inspect stored grounded reasoning trajectories.

Arguments:

  • trajectory_id

Flags:

  • --task-family — Filter by task family
  • --limit — Number of trajectories to list
  • --json — Output JSON

sb grounded report

Show score trends and aggregate metrics across grounded reasoning runs.

Flags:

  • --last, -n — Number of recent trajectories to include
  • --family — Filter by task family
  • --json — Output JSON

sb grounded synth

Generate synthetic grounded QA dataset from corpus.

Arguments:

  • seeds

Flags:

  • --output — Output JSONL path
  • --provider — LLM provider name
  • --model — Model name
  • --top-k — Chunks per seed retrieval

sb habits

Habits tracker — daily checklist from vault/HABITS.md.

sb habits check

Mark a habit as complete for today.

Arguments:

  • name (required)

sb habits nudge

Get an LLM-generated motivational nudge for incomplete habits.

Flags:

  • --no-llm — Skip LLM, print pending habits only

sb habits show

Show today's habit checklist with completion status.

sb habits template

Create a starter vault/HABITS.md template.

sb harness

Harness Router — unified interface for Claude Code, Codex, and Gemini CLI with auto-selection.

sb harness audit

Grade this runtime against the 8-outcome / 40-control harness audit.

Flags:

  • --json — Emit the full audit as JSON
  • --fail-under — Exit non-zero if the overall score is below this (0..1); use in CI gates.

sb harness check

Quick health check — show which harnesses are ready.

sb harness list

List all known harnesses and their availability status.

Flags:

  • --json — Emit JSON
  • --extended, -x — Include probe-only CLIs (cursor-agent, opencode, pi, openclaw) that are detected but not yet dispatchable.

sb harness review

Review a file/directory using the best available harness.

Arguments:

  • path (required)

Flags:

  • --harness, -H — Force a specific harness: claude | codex (default: claude)
  • --context, -c — Topic for vault context lookup
  • --model, -m — Model override
  • --cwd — Working directory
  • --add-dir — Extra directories to expose (repeatable)
  • --no-context — Skip vault context injection
  • --approval-mode — Codex-only: suggest | auto-edit | full-auto
  • --permission-mode — Claude-only: permission mode override
  • --json — Emit JSON result

sb harness run

Route a task to the best available harness (Claude Code or Codex).

Arguments:

  • task (required)

Flags:

  • --harness, -H — Force a specific harness: claude | codex (default: auto-select)
  • --context, -c — Topic to fetch from vault for context (default: task text)
  • --model, -m — Model override passed to the selected harness
  • --cwd — Working directory
  • --add-dir — Extra directories to expose (repeatable)
  • --top-k — Vault chunks to inject
  • --no-context — Skip vault context injection
  • --approval-mode — Codex-only: suggest | auto-edit | full-auto
  • --permission-mode — Claude-only: permission mode override
  • --json — Emit JSON result

sb heartbeat

Heartbeat — daily briefing from all connectors + LLM synthesis.

sb heartbeat install

Install the heartbeat as a macOS LaunchAgent.

Flags:

  • --hour — Hour to run (0-23)
  • --minute — Minute to run (0-59)

sb heartbeat run

Run the heartbeat: gather + synthesize + write daily log.

Flags:

  • --dry-run — Gather data but skip LLM synthesis
  • --no-state — Skip state diffing; always run
  • --quiet, -q — Suppress output

sb heartbeat status

Show heartbeat status — last run, next scheduled run.

sb heartbeat uninstall

Uninstall the heartbeat LaunchAgent.

sb home

Open the guided home screen with common next actions.

sb hooks

Claude Code hooks — auto-inject SecondBrain memory into every session.

sb hooks install

Install SecondBrain hooks into ~/.claude/settings.json.

Flags:

  • --dry-run — Preview changes without writing
  • --force — Overwrite existing hook scripts

sb hooks show

Show the current Claude Code hooks configuration.

sb hooks uninstall

Remove SecondBrain hooks from ~/.claude/settings.json.

sb host

Per-machine identity (machine_id, label) for fleet coordination.

sb host get

Print a single identity field (handy for shell pipelines).

Arguments:

  • field

sb host rename

Set the human-readable label for this machine.

Arguments:

  • label (required)

sb host show

Show this machine's identity (creates one on first call).

Flags:

  • --json — Emit JSON.

sb identity

Semantic identity and lineage operations.

Arguments:

  • from_semantic_id (required)
  • to_semantic_id (required)

Flags:

  • --relation

sb identity mint

Arguments:

  • namespace (required)
  • entity_type (required)
  • label (required)

Flags:

  • --internal-type
  • --internal-id
  • --version

sb identity neighbors

Arguments:

  • semantic_id (required)

Flags:

  • --limit

sb identity parse

Arguments:

  • value (required)

sb identity people-candidate-decide

Arguments:

  • candidate_id (required)

Flags:

  • --decision — approved|rejected|open
  • --matched-person-id — Optional canonical person_id
  • --notes — Optional review notes
  • --decided-by — Optional reviewer identity
  • --json — Output JSON

sb identity people-candidates

Flags:

  • --status — Optional candidate status filter
  • --kind — Optional candidate kind filter
  • --source-kind — Optional source kind filter
  • --limit
  • --json — Output JSON

sb identity people-prune

Flags:

  • --apply — Delete the matched rows
  • --limit
  • --json — Output JSON

sb identity people-reconcile

Flags:

  • --limit
  • --json — Output JSON

Arguments:

  • query

Flags:

  • --limit
  • --json — Output JSON

sb identity people-sync

Flags:

  • --source — all|tasks|meetings|apple
  • --limit
  • --json — Output JSON

sb identity person-observations

Arguments:

  • person_id (required)

Flags:

  • --type — Optional observation type filter
  • --source-kind — Optional source kind filter
  • --limit
  • --json — Output JSON

sb identity person-relationships

Arguments:

  • person_id (required)

Flags:

  • --type — Optional relationship type filter
  • --target-type — Optional target entity type filter
  • --source-kind — Optional source kind filter
  • --limit
  • --json — Output JSON

sb identity person-show

Arguments:

  • person_id (required)

Flags:

  • --json — Output JSON

sb identity unresolved-owners

Flags:

  • --limit
  • --json — Output JSON

sb improve

Continuously improve conversation quality from real failures.

sb improve conversations

Intake, review, and validate continuous conversation improvements.

sb improve conversations decide

Record human decisions for one or more improvement cases.

Arguments:

  • batch_id (required)

Flags:

  • --case — Improvement case id
  • --decision — approved|deferred|rejected
  • --approve-all-runnable — Approve every auto-runnable non-guard case in the batch
  • --notes — Optional operator notes
  • --json — Output JSON

sb improve conversations intake

Create a new conversation-improvement batch from recent failures and guards.

Flags:

  • --recent-sessions
  • --recent-events
  • --synthetic-guard
  • --synthetic-pack
  • --json — Output JSON

sb improve conversations report

Show the latest improvement run report for one batch or globally.

Arguments:

  • batch_id

Flags:

  • --latest — Show the latest report
  • --json — Output JSON

sb improve conversations review

Show prioritized cases and current decisions for one batch.

Arguments:

  • batch_id (required)

Flags:

  • --json — Output JSON

sb improve conversations run

Run approved improvements and validate them against replays and guards.

Arguments:

  • batch_id (required)

Flags:

  • --json — Output JSON

sb improve conversations session

Run the session-scoped conversation improvement loop for one chat session.

Arguments:

  • session_id (required)

Flags:

  • --provider — Judge provider name
  • --attempts — Override the lane attempt cap
  • --metric-threshold — Deterministic score threshold on a 0-100 scale
  • --judge-threshold — LLM judge threshold on a 0-100 scale
  • --synthetic-guard — Synthetic guard suite to run alongside the selected session
  • --synthetic-pack — Synthetic guard pack
  • --json — Output JSON

sb inbox

Inbox briefing across email and calendar context.

sb inbox brief

Summarize inbox and calendar context in one operator-facing brief.

Flags:

  • --query, -q — Gmail search query
  • --max-messages — Maximum email items to summarize
  • --days — Meeting horizon in days
  • --max-events — Maximum events to summarize
  • --all — Include all configured Google calendars
  • --json — Output raw JSON

sb inbox promote-commitments

Promote conversation-derived email promises into durable commitments.

Flags:

  • --query, -q — Gmail search query
  • --max-messages — Maximum email items to inspect
  • --queue-reminders — Immediately queue governed reminder approvals for due-soon promoted commitments.
  • --reminder-days-ahead — Only queue reminders for promoted commitments due within this many days.
  • --dry-run — Preview derived promotions without creating commitments
  • --json — Output raw JSON

sb index

Alias for sb context index (optional offline index build).

sb indriya

Indriya — operational local-first source runtime and control plane.

sb indriya disable

Disable one Indriya source.

Arguments:

  • source (required)

Flags:

  • --json — Output JSON

sb indriya doctor

Diagnose stale or failing Indriya sources.

Flags:

  • --stale-after-hours
  • --json — Output JSON

sb indriya enable

Enable one Indriya source.

Arguments:

  • source (required)

Flags:

  • --path — Override source path(s)
  • --host — Browser clip host override
  • --port — Browser clip port override
  • --endpoint — Browser clip localhost endpoint override
  • --json — Output JSON

sb indriya health

Show Indriya runtime health and source state.

Flags:

  • --json — Output JSON

sb indriya insights

Summarize recent Indriya context into a compact personal snapshot.

Flags:

  • --hours — Insight window
  • --limit — Max captures to analyze
  • --source — Only include selected source ids
  • --json — Output JSON

sb indriya recent

Show recent Indriya-backed captures and source aggregates.

Flags:

  • --hours — Recent context window
  • --limit — Max captures to show
  • --source — Only include selected source ids
  • --json — Output JSON

sb indriya run

Run enabled Indriya sources once.

Flags:

  • --source — Only run selected source ids
  • --json — Output JSON

sb indriya sources

List built-in Indriya sources and their current state.

Flags:

  • --json — Output JSON

sb indriya watch

Watch enabled Indriya sources continuously.

Flags:

  • --source — Only watch selected source ids
  • --interval-seconds — Polling interval override

sb ingest

Ingest markdown/pdf/text files into the local knowledge store.

sb ingest-add

Add raw artifacts to the context refinery without promoting them yet.

sb ingest-inspect

Inspect one context-refinery artifact with segments, scores, and promotions.

sb ingest-normalize

Normalize raw artifacts into semantic segments.

sb ingest-pipeline

Run add -> normalize -> score -> promote for a path.

sb ingest-promote

Promote eligible scored segments into typed knowledge stores.

sb ingest-quarantine

Inspect likely low-signal or quarantine-worthy segments.

sb ingest-refresh

Refresh all registered ingest sources (incremental by default).

sb ingest-repo

Clone a git repository and ingest its documentation files.

sb ingest-reprocess

Reprocess an existing corpus through normalize -> score -> promote.

sb ingest-score

Score normalized segments for searchability and promotion.

sb ingest-source-disable

Disable a source (skip during refresh without removing it).

sb ingest-source-enable

Re-enable a previously disabled source.

sb ingest-source-interval

Change the refresh interval for a source.

sb ingest-source-remove

Remove a path from the ingest source registry.

sb ingest-sources

List all registered ingest sources and their staleness.

sb ingest-stats

Show context-refinery artifact and promotion stats.

sb integrations

Integration registry and sync automation operations.

sb integrations add-apple-flagged-mail

Arguments:

  • name (required)

Flags:

  • --account — Optional Apple Mail account filter.
  • --mailbox — Optional Apple Mail mailbox filter.
  • --days — Only consider flagged mail from this many recent days.
  • --max-results — Maximum flagged messages to process per run.
  • --only-new, --all — By default only process flagged messages not seen in prior successful runs.
  • --project — Optional project hint for resulting captures.
  • --tag — Tags applied to resulting captures (repeatable).
  • --interval-seconds — Target sync cadence for due-based runs.
  • --paused — Create integration paused.
  • --json

sb integrations add-apple-next-meeting

Arguments:

  • name (required)

Flags:

  • --project — Optional project hint for resulting captures.
  • --tag — Tags applied to resulting captures (repeatable).
  • --interval-seconds — Target sync cadence for due-based runs.
  • --paused — Create integration paused.
  • --json

sb integrations add-github

Arguments:

  • name (required)

Flags:

  • --repo — GitHub repo slug like owner/repo
  • --events-file — JSONL file with GitHub event payloads (absolute or relative to vault).
  • --output-dir — Directory where normalized markdown artifacts are written.
  • --interval-seconds — Target sync cadence for due-based runs.
  • --paused — Create integration paused.
  • --json

sb integrations add-inbox

Arguments:

  • name (required)

Flags:

  • --path — Inbox path (absolute or relative to vault)
  • --include-glob — Optional glob filter within inbox path (repeatable).
  • --recursive, --no-recursive
  • --interval-seconds — Target sync cadence for due-based runs.
  • --paused — Create integration paused.
  • --json

sb integrations add-mcp

Arguments:

  • name (required)

Flags:

  • --server — Configured MCP server name.
  • --tool — MCP tool name to call on sync.
  • --args-json — JSON object of MCP tool arguments.
  • --record-path — Dot-path to list of records in MCP output.
  • --id-field — Record field/path for stable id.
  • --title-field — Record field/path for markdown title.
  • --body-field — Record field/path for markdown body.
  • --updated-at-field
  • --url-field
  • --cursor-arg — MCP argument name that receives cursor value.
  • --cursor-field — Dot-path in MCP output for next cursor value.
  • --output-dir — Directory where normalized markdown artifacts are written.
  • --interval-seconds — Target sync cadence for due-based runs.
  • --paused — Create integration paused.
  • --json

sb integrations add-mcp-preset

Arguments:

  • name (required)

Flags:

  • --preset — Curated MCP preset. Supported values: atlassian, github, m365_calendar
  • --tool — MCP tool name to call on sync.
  • --args-json — JSON object of MCP tool arguments.
  • --record-path — Override preset record path.
  • --id-field — Override preset record id field.
  • --title-field — Override preset title field.
  • --body-field — Override preset body field.
  • --updated-at-field — Override preset updated-at field.
  • --url-field — Override preset URL field.
  • --cursor-arg — Override preset cursor argument.
  • --cursor-field — Override preset cursor field.
  • --output-dir — Override preset artifact root.
  • --interval-seconds — Target sync cadence for due-based runs.
  • --paused — Create integration paused.
  • --json

sb integrations add-outlook-mail

Arguments:

  • name (required)

Flags:

  • --server — Configured MCP server name for Outlook mail search.
  • --query — Search query passed to the Outlook mail MCP server.
  • --category — Require one or more Outlook categories on matched messages (repeatable).
  • --flagged-only — Keep only flagged messages after retrieval.
  • --max-results — Maximum messages to retrieve before local filters are applied.
  • --output-dir — Directory where normalized markdown artifacts are written.
  • --interval-seconds — Target sync cadence for due-based runs.
  • --paused — Create integration paused.
  • --json

sb integrations daemon

Flags:

  • --interval-seconds
  • --once
  • --max-cycles — 0 means run until interrupted.
  • --dry-run
  • --force
  • --due-only, --all — Run only due integrations (default).
  • --type
  • --limit
  • --pid-file — PID lock file path.
  • --json

sb integrations github-backfill

Arguments:

  • identifier (required)

Flags:

  • --kind — issues|pulls|both
  • --state — open|closed|all
  • --limit — Max records per resource to fetch.
  • --since — Optional ISO timestamp floor (updated_at).
  • --token-env — Env var containing GitHub token.
  • --sync-after-backfill, --no-sync-after-backfill
  • --force-sync
  • --dry-run
  • --json

sb integrations github-poll

Arguments:

  • identifier (required)

Flags:

  • --limit — Max events to fetch from GitHub API (1..100).
  • --token-env — Env var containing GitHub token.
  • --sync-after-poll, --no-sync-after-poll
  • --force-sync
  • --dry-run
  • --json

sb integrations github-webhook

Arguments:

  • identifier (required)

Flags:

  • --payload-file — Webhook JSON payload file path or '-' for stdin.
  • --event-type — X-GitHub-Event value.
  • --delivery-id — X-GitHub-Delivery value.
  • --signature — X-Hub-Signature-256 header value.
  • --secret-env
  • --sync-after-ingest, --no-sync-after-ingest
  • --force-sync
  • --dry-run
  • --json

sb integrations health

Flags:

  • --type
  • --limit
  • --max-stale-24h — Fail if stale_24h_count exceeds this value.
  • --max-failing — Fail if failing_count exceeds this value.
  • --json

sb integrations list

Flags:

  • --type
  • --status
  • --limit
  • --json

sb integrations pause

Arguments:

  • identifier (required)

Flags:

  • --json

sb integrations reset-cursor

Arguments:

  • identifier (required)

Flags:

  • --line — Line cursor to resume from (default: 0)
  • --json

sb integrations resume

Arguments:

  • identifier (required)

Flags:

  • --json

sb integrations run-due

Flags:

  • --dry-run
  • --force
  • --type
  • --limit
  • --json

sb integrations runs

Arguments:

  • identifier (required)

Flags:

  • --limit
  • --json

sb integrations show

Arguments:

  • identifier (required)

Flags:

  • --json

sb integrations sync

Arguments:

  • identifier (required)

Flags:

  • --dry-run
  • --force
  • --json

sb integrations sync-all

Flags:

  • --dry-run
  • --force
  • --due-only — Run only integrations that are due by cadence.
  • --type
  • --limit
  • --json

sb integrations test

Arguments:

  • identifier (required)

Flags:

  • --force — Pass through force flag to dry-run sync probe.
  • --skip-sync — Only run config/source checks (no sync probe).
  • --json

sb integrations watch

Flags:

  • --interval-seconds
  • --once
  • --max-cycles — 0 means run until interrupted.
  • --dry-run
  • --force
  • --due-only, --all — Run only due integrations (default).
  • --type
  • --limit
  • --json

sb kernel

SecondBrain Kernel — core OS-layer for all agentic AI (RunContext, ToolRegistry, Memory).

sb kernel hooks

List registered pre/post tool hooks.

Flags:

  • --json — Output JSON.

sb kernel info

Show kernel package overview and feature inventory.

Flags:

  • --json — Output JSON.

sb kernel namespaces

List all registered decision namespaces and their keys.

Flags:

  • --json — Output JSON.

sb kernel tools

List tools available via the MeetingCopilot kernel registry.

Flags:

  • --agent, -a — Filter tools by agent allowlist.
  • --safety, -s — Max safety class (read_only|local_write|network|destructive).
  • --json — Output JSON.

sb keys

Keys — OS keychain management for provider API keys (store, retrieve, delete).

sb keys delete

Remove a key from the OS keychain.

Arguments:

  • name (required)

Flags:

  • --yes, -y — Skip confirmation

sb keys get

Show the stored value for a key (masked by default).

Arguments:

  • name (required)

Flags:

  • --reveal — Show full value (not masked)

sb keys list

List key names stored via SecondBrain (values are never shown).

Flags:

  • --json — Emit JSON

sb keys set

Store an API key in the OS keychain.

Arguments:

  • name (required)
  • value

sb keys status

Show all known provider keys and where each is loaded from.

Flags:

  • --json — Emit JSON

sb knowledge

Promoted knowledge inspection operations.

sb knowledge inspect

Inspect promoted facts, workflows, decisions, and open loops with provenance.

Arguments:

  • id_or_query (required)

Flags:

  • --json — Output JSON

sb knowledge manan

Manan: reflection, questions, and synthesis.

sb knowledge manan reflect

Reflect on Shravan intake before it can mature into practice.

Arguments:

  • id_or_capture_id (required)

Flags:

  • --question — Additional question to resolve
  • --connect — Existing memory, project, or concept connection
  • --counterpoint — Counterpoint to include
  • --resolved — Resolved understanding override
  • --json — Output JSON

sb knowledge nididhyasan

Nididhyasan: convert reflection into practice.

sb knowledge nididhyasan implement

Convert Manan reflection into a practice loop or action rule.

Arguments:

  • manan_id (required)

Flags:

  • --type — Implementation type
  • --title, -t — Action title
  • --owner — Action owner
  • --frequency — Practice cadence
  • --review-prompt — Review prompt for the practice loop
  • --success-metric — How to know this worked
  • --status — Initial action status
  • --json — Output JSON

sb knowledge review

Review captured knowledge that has not yet become reflected or practiced.

Flags:

  • --weekly — Render weekly review prompts
  • --limit, -n — Maximum records per bucket
  • --json — Output JSON

sb knowledge shravan

Shravan: source-aware knowledge intake.

sb knowledge shravan add

Capture Shravan intake with source authority, provenance, and questions.

Arguments:

  • content

Flags:

  • --file, -f — Read intake from file
  • --capture-id — Build intake from an existing capture id
  • --source — Source type
  • --title, -t — Source title
  • --authority — Source authority level
  • --lineage — Source lineage
  • --source-ref — Source reference
  • --trust-score — Reliability score
  • --freshness-score — Freshness score
  • --relevance-score — Relevance score
  • --why — Why this intake matters
  • --open-question — Open question to carry into Manan
  • --json — Output JSON

sb knowledge status

Show knowledge assimilation counts and maturity states.

Flags:

  • --limit, -n — Number of recent items
  • --json — Output JSON

sb linear

Linear — read issues and projects.

sb linear auth

Save your Linear API key to keys.env.

sb linear ingest

Ingest Linear issues into SecondBrain memory.

Flags:

  • --states — Comma-separated state names to ingest
  • --limit, -n

sb linear issues

List Linear issues.

Flags:

  • --states — Comma-separated state names to filter
  • --me — Only issues assigned to me
  • --team — Filter by team key
  • --limit, -n
  • --json

sb linear projects

List Linear projects.

Flags:

  • --limit, -n
  • --json

sb linear teams

List Linear teams.

sb listen

Capture live call audio and extract meeting intelligence.

sb listen devices

List available audio input devices to find the BlackHole 2ch ID.

sb listen start

Start live audio capture. Press Ctrl-C to stop and extract intelligence.

Flags:

  • --title, -t — Meeting title.
  • --device, -d — BlackHole device ID (overrides config.toml).
  • --chunk — Chunk size in seconds (default: 30).
  • --model — MLX Whisper model repo (default: whisper-base-mlx).
  • --lang — Spoken language code.
  • --provider — LLM provider for extraction (anthropic, openai, …). Empty = auto.
  • --ingest, --no-ingest — Ingest the vault note into the vector store after capture.
  • --json — Emit JSON output.

sb maintenance

Maintenance runtime — manual ticks + config for the periodic housekeeping loop.

sb maintenance config

Show the effective maintenance configuration.

sb maintenance tick

Run a single maintenance tick synchronously (no background loop).

Flags:

  • --json — Emit JSON.
  • --archive-ttl-hours — Drop bridge-archive envelopes older than N hours.

sb marketer

marketer — end-to-end campaign planning agent.

sb marketer demo

Run 3 canned demo campaigns and show results.

Flags:

  • --seed — Random seed for reproducible demo

sb marketer eval

Evaluate a marketing brief or existing campaign.

Arguments:

  • brief_or_id (required)

Flags:

  • --simulate, --no-simulate — Run simulation
  • --seed — Random seed
  • --json — Output raw JSON

sb marketer plan

Plan a marketing campaign end-to-end from a 2–6 line brief.

Arguments:

  • brief (required)

Flags:

  • --dry-run, --no-dry-run — Dry run (no real sends)
  • --out, -o — Extra output path for campaign_spec.json
  • --seed — Random seed for reproducible simulation
  • --json — Output raw JSON

sb mcp

MCP registry, discovery, and execution operations.

sb mcp adapters

Show normalized adapter inventory across MCP, GitHub, Codex, and artifact tools.

Flags:

  • --json — Output JSON

sb mcp approvals

Inspect and manage remembered MCP approvals.

sb mcp approvals explain

Explain one remembered MCP approval.

Arguments:

  • memory_id (required)

Flags:

  • --json — Output JSON

sb mcp approvals list

List remembered MCP approvals.

Flags:

  • --session-id — Optional chat session id
  • --json — Output JSON

sb mcp approvals revoke

Revoke one remembered MCP approval.

Arguments:

  • memory_id (required)

Flags:

  • --reason — Revocation reason
  • --json — Output JSON

sb mcp ban-capability

Ban one capability from planner exposure and execution.

Arguments:

  • server_id (required)
  • capability_name (required)

Flags:

  • --reason — Ban reason
  • --json — Output JSON

sb mcp call

Call one approved MCP tool through policy and tracing layers.

Arguments:

  • server_id (required)
  • capability_name (required)

Flags:

  • --json-input, --args-json — JSON object tool args
  • --confirm — Allow side-effecting tools
  • --json — Output JSON

sb mcp compile-resource

Fetch one MCP resource and compile it into durable local context.

Arguments:

  • server_id (required)
  • resource_name (required)

Flags:

  • --json — Output JSON

sb mcp discover

Discover normalized capabilities for one approved MCP server.

Arguments:

  • server_id (required)

Flags:

  • --refresh — Bypass in-process discovery cache
  • --intent — Intent text used for capability narrowing
  • --agent-profile — Agent profile used for capability narrowing
  • --limit — Maximum capabilities to show
  • --json — Output JSON

sb mcp doctor

Validate MCP config and check enabled server reachability.

Flags:

  • --deep — Include score and approval-memory diagnostics
  • --json — Output JSON

sb mcp eval

Run and inspect MCP effectiveness evals.

sb mcp eval report

Show the latest MCP eval report.

Flags:

  • --suite — Optional suite name
  • --json — Output JSON

sb mcp eval run

Run MCP effectiveness evals.

Arguments:

  • suite

Flags:

  • --scenario-file — JSON scenario file
  • --ablate-capability — Qualified capability to ablate
  • --ablate-server — Server id to ablate
  • --disable-approval-memory — Ignore approval-memory effects during eval
  • --json — Output JSON

sb mcp explain

Explain normalized metadata, policy, and exposure state for one capability.

Arguments:

  • server_id (required)
  • capability_name (required)

Flags:

  • --kind — tool|resource|prompt
  • --intent — Intent used for exposure routing
  • --agent-profile — Agent profile used for exposure routing
  • --json — Output JSON

sb mcp exposure

Explain adaptive MCP capability exposure.

sb mcp exposure explain

Show which capabilities would be exposed and why.

Flags:

  • --task — Task text to evaluate
  • --agent-profile — Agent profile for routing
  • --limit — Maximum exposed capabilities
  • --json — Output JSON

sb mcp feedback

Record and inspect MCP task outcome feedback.

sb mcp feedback recent

Show recent MCP task outcome feedback.

Flags:

  • --limit — Maximum feedback rows
  • --json — Output JSON

sb mcp feedback record

Record MCP task outcome feedback and attribute it to used capabilities.

Flags:

  • --session-id — Chat/session id to attribute
  • --task — Task summary
  • --success, --failure — Whether the task succeeded
  • --notes — Free-form notes
  • --capabilities — Comma-separated qualified capability list
  • --json — Output JSON

sb mcp marketplace

Browse and configure community MCP servers.

sb mcp marketplace configure

Print a SecondBrain MCP config snippet for the given server.

Arguments:

  • server_id (required)

Flags:

  • --json — Emit JSON config snippet

sb mcp marketplace info

Show full details and install instructions for an MCP server.

Arguments:

  • server_id (required)

Flags:

  • --json — Output JSON

sb mcp marketplace list

Browse the curated SecondBrain MCP marketplace.

Flags:

  • --category, -c — Filter by category
  • --trust, -t — Filter by trust level
  • --json — Output JSON

Search the MCP marketplace by keyword.

Arguments:

  • query (required)

Flags:

  • --json — Output JSON

sb mcp pin-capability

Pin one capability so exposure prefers it.

Arguments:

  • server_id (required)
  • capability_name (required)

Flags:

  • --reason — Pin reason
  • --json — Output JSON

sb mcp quarantine

Manually quarantine one MCP server.

Arguments:

  • server_id (required)

Flags:

  • --reason — Quarantine reason
  • --json — Output JSON

sb mcp read-resource

Read one approved MCP resource and optionally ingest it into local context.

Arguments:

  • server_id (required)
  • resource_name (required)

Flags:

  • --store — Persist and ingest the resource into local context
  • --json — Output JSON

sb mcp report

Materialize and inspect MCP intelligence snapshots.

sb mcp report latest

Show the latest materialized MCP intelligence snapshot paths.

Flags:

  • --json — Output JSON

sb mcp report snapshot

Build an MCP intelligence snapshot report.

Flags:

  • --write, --no-write — Persist snapshot files under state/mcp_reports
  • --json — Output JSON

sb mcp score

Show capability and server scoreboards.

sb mcp score capabilities

Show MCP capability quality scores.

Flags:

  • --json — Output JSON

sb mcp score servers

Show MCP server reputation scores.

Flags:

  • --json — Output JSON

sb mcp servers

List configured MCP servers and their policy/compatibility posture.

Flags:

  • --enabled-only — Show only enabled servers
  • --json — Output JSON

sb mcp unban-capability

Remove a manual ban from one capability.

Arguments:

  • server_id (required)
  • capability_name (required)

Flags:

  • --json — Output JSON

sb mcp unquarantine

Remove a manual quarantine from one MCP server.

Arguments:

  • server_id (required)

Flags:

  • --json — Output JSON

sb mcp warmup

Warm enabled MCP servers by running discovery and health checks.

Flags:

  • --refresh, --no-refresh — Refresh discovery during warmup
  • --json — Output JSON

sb meet

Capture meeting notes/transcript with structured extraction.

sb meeting

MeetingCopilot — transcript to action items.

sb meeting close

Close a meeting into persisted decisions, commitments, and follow-ups.

Arguments:

  • transcript_file (required)

Flags:

  • --project — Optional project label
  • --project-id — Optional project id for created follow-up tasks
  • --meeting-title — Meeting title shown in generated artifacts
  • --meeting-date — Meeting date (YYYY-MM-DD). Defaults to current date.
  • --engine — Workflow engine: auto|local|adk
  • --adk-model — ADK model when --engine adk (or auto with ADK available).
  • --ping-owners, --no-ping-owners — Generate owner ping draft file grouped by assignee
  • --a2a-pings, --no-a2a-pings — Also publish owner pings to the local A2A queue
  • --json — Output JSON.

sb meeting extract

Extract action items, risks, and follow-ups from a meeting transcript.

Arguments:

  • transcript_file

Flags:

  • --text, -t — Transcript as inline string.
  • --title — Meeting title.
  • --participants, -p — Comma-separated participant names.
  • --json — Output JSON.
  • --provider — LLM provider name (anthropic, openai, …). Empty = chain default.
  • --model — Model name override for the chosen provider.

sb meeting prep

Prepare for an upcoming meeting using calendar, email, and project context.

Arguments:

  • selector

Flags:

  • --project — Optional project label for continuity context
  • --project-id — Optional project id for linked tasks and continuity
  • --days — Calendar horizon in days
  • --mail-days — How far back to search attendee email
  • --max-emails — Maximum attendee email items to include
  • --all — Include all configured Google calendars
  • --json — Output JSON.

sb memory

Persistent memory subsystem operations.

sb memory add

Arguments:

  • content (required)

Flags:

  • --category
  • --importance
  • --pin
  • --status
  • --source-type
  • --confidence
  • --provenance
  • --evidence
  • --json

sb memory approve

Arguments:

  • memory_id (required)

Flags:

  • --json

sb memory benchmark

Memory benchmark generation and evaluation.

sb memory benchmark datasets

Download public memory benchmark datasets.

sb memory benchmark datasets download

Download public memory benchmark datasets into a local directory.

Flags:

  • --suite
  • --variant
  • --out
  • --include-large
  • --force
  • --json

sb memory benchmark datasets list

List known public memory benchmark dataset download targets.

Flags:

  • --include-large
  • --json

sb memory benchmark generate

Generate an offline-safe synthetic memory benchmark fixture.

Flags:

  • --suite
  • --style
  • --cases
  • --out
  • --json

sb memory benchmark report

Show a persisted memory benchmark run.

Flags:

  • --run-id
  • --json

sb memory benchmark run

Run a memory benchmark in an isolated temp memory store.

Flags:

  • --suite — Benchmark suite: auto, synthetic, locomo, longmemeval, or beam.
  • --dataset-path
  • --top-k
  • --judge-policy
  • --max-cases
  • --cost-model
  • --token-budget
  • --json

sb memory consolidate

Flags:

  • --session-id
  • --json

sb memory context

Arguments:

  • query (required)

Flags:

  • --session-id
  • --top-k
  • --debug
  • --json

sb memory decay

Flags:

  • --older-than-days
  • --dry-run
  • --json

sb memory dedup

Flags:

  • --dry-run
  • --json

sb memory delete

Arguments:

  • memory_id (required)

Flags:

  • --confirm — Required confirmation flag
  • --json

sb memory edit

Arguments:

  • memory_id (required)

Flags:

  • --content
  • --category
  • --status
  • --importance
  • --json

sb memory episodes

Flags:

  • --session-id
  • --since
  • --until
  • --limit
  • --json

sb memory evals

Run deterministic governance evals over the memory flywheel.

Flags:

  • --limit
  • --stale-confidence-threshold
  • --json

sb memory events

Flags:

  • --episode-id
  • --limit
  • --json

sb memory evidence

Source evidence recall index operations.

sb memory evidence reindex

Flags:

  • --limit
  • --json

Arguments:

  • query (required)

Flags:

  • --top-k
  • --source-type
  • --source-ref
  • --json

sb memory evidence show

Arguments:

  • evidence_id (required)

Flags:

  • --json

sb memory evidence stats

Flags:

  • --json

sb memory export

Flags:

  • --file
  • --format
  • --category
  • --status
  • --include-working
  • --include-episodes
  • --limit
  • --json

sb memory import

Arguments:

  • file (required)

Flags:

  • --mode
  • --dry-run
  • --json

sb memory list

Flags:

  • --category
  • --status
  • --sort
  • --limit
  • --json

sb memory maintenance

Flags:

  • --dry-run
  • --lookback-days
  • --min-repeats
  • --older-than-days
  • --json

sb memory merge

Arguments:

  • memory_a_id (required)
  • memory_b_id (required)

Flags:

  • --json

sb memory migrate

Migrate vectors from one backend to another (e.g. Chroma → LanceDB).

Flags:

  • --from, -f — Source backend (chroma|lancedb|sqlite_vec|qdrant|memory)
  • --to, -t — Target backend
  • --from-dir — Source persist dir (default: state/chroma for chroma, else state/retrieval)
  • --to-dir — Target persist dir (default: state/retrieval)
  • --collections — Comma-separated collection names to migrate
  • --dry-run — Report counts without writing
  • --json

sb memory pin

Arguments:

  • memory_id (required)

Flags:

  • --json

sb memory promote

Run one promotion pass: working → episodic, episodic → durable, decay stale.

Flags:

  • --dry-run — Evaluate without writing.
  • --json
  • --verbose, -v

sb memory review-add

Arguments:

  • statement (required)

Flags:

  • --memory-type
  • --confidence
  • --ttl
  • --scope
  • --requires-approval, --no-requires-approval
  • --provenance
  • --task-id
  • --json

sb memory review-apply

Arguments:

  • proposal_id (required)

Flags:

  • --json

sb memory review-approve

Arguments:

  • proposal_id (required)

Flags:

  • --approver
  • --rationale
  • --apply
  • --json

sb memory review-list

Flags:

  • --status
  • --trace-id — Filter by source trace id
  • --task-id — Filter by task id
  • --limit
  • --json

sb memory review-reject

Arguments:

  • proposal_id (required)

Flags:

  • --approver
  • --rationale
  • --json

sb memory review-show

Arguments:

  • proposal_id (required)

Flags:

  • --json

sb memory review-summary

Summarize memory proposals, promotion evidence, and applied memory IDs.

Flags:

  • --status
  • --limit
  • --json

Arguments:

  • query (required)

Flags:

  • --top-k
  • --category
  • --status
  • --json

sb memory show

Arguments:

  • memory_id (required)

Flags:

  • --json

sb memory stats

Flags:

  • --json

sb memory synthesize

Flags:

  • --lookback-days
  • --min-repeats
  • --dry-run
  • --json

sb memory timeline

Show a chronological timeline of memories grouped by day.

Flags:

  • --days — How many days back to show.
  • --limit — Max memories to load.
  • --kind — Filter by memory kind (e.g. long_term, episode).
  • --json — Emit JSON instead of rich table.

sb memory unpin

Arguments:

  • memory_id (required)

Flags:

  • --json

sb memory working

Flags:

  • --session-id
  • --json

sb models

List available models for a provider using provider APIs where possible.

sb music

Manage local-first music preferences.

sb music connect

Connect a music provider (e.g. Spotify) via interactive OAuth flow.

Arguments:

  • provider (required)

sb music debug-profile

Print the raw derived profile snapshot for a provider.

Arguments:

  • provider

sb music history

Show recent listening history.

Flags:

  • --provider — Provider to query
  • --limit — Number of recent events to show

Link an existing music profile snapshot to a person in the people graph.

Arguments:

  • person_id (required)

Flags:

  • --provider — Provider whose profile to link.

sb music podcasts

Show saved podcast shows and recent episodes.

Flags:

  • --provider — Provider to query
  • --limit — Max episodes to show

sb music purge

Delete all synced music data for a provider from local storage.

Arguments:

  • provider

sb music recommend

Recommend a track grounded in synced music preferences.

Arguments:

  • mood

Flags:

  • --provider — Provider to use

sb music status

Show the currently connected providers, sync status, and stored entity counts.

sb music sync

Sync listening history, top items, and derived profile from connected providers.

Arguments:

  • provider

Flags:

  • --person-id — Link music profile to a person in the people graph.

sb music top

Show your top tracks or artists from the latest profile snapshot.

Arguments:

  • kind

Flags:

  • --provider — Provider to query
  • --limit — Number of items to show

sb n8n

n8n workflow automation — invoke governed workflows as tools.

sb n8n list

List all workflows available on the n8n instance.

Flags:

  • --json — Output JSON

sb n8n registered

Show n8n workflows registered as SecondBrain tools.

Flags:

  • --json — Output JSON

sb n8n run

Invoke an n8n workflow through ToolExecutor and ToolPolicy.

Arguments:

  • workflow_id (required)

Flags:

  • --input, -i — JSON input payload
  • --path — Override webhook path for an ad hoc workflow
  • --json — Output JSON result

sb n8n status

Check n8n daemon connectivity and configuration.

Flags:

  • --json — Output JSON status

sb note

Quick-capture a note into SecondBrain.

sb notifications

Typed notifications — first-party SB inbox (session completions, decisions, autopilot runs).

sb notifications archive

Archive notifications (removes from default list views).

Arguments:

  • notification_ids (required)

sb notifications clear

Delete notifications permanently.

Flags:

  • --archived-only, --all — By default only archived are deleted. Pass --all to wipe everything.
  • --yes, -y — Skip confirmation

sb notifications emit

Emit a notification (handy for testing hooks).

Arguments:

  • type (required)

Flags:

  • --subject, -s — Short subject line
  • --body, -b — Longer detail body
  • --severity — info | attention | action_required
  • --agent — Target agent slug (default: user)

sb notifications list

List notifications (defaults to unarchived across all recipients).

Flags:

  • --recipient, -r — Filter by recipient type: user | agent
  • --agent — Filter to a specific agent slug (implies --recipient agent)
  • --unread, -u — Only show unread notifications
  • --all, -a — Include archived notifications
  • --severity, -s — Filter by severity: info | attention | action_required
  • --type, -t — Filter by event type (e.g. session.completed)
  • --limit, -n — Max rows
  • --json — Emit JSON

sb notifications read

Mark notifications as read.

Arguments:

  • notification_ids (required)

Flags:

  • --all — Mark ALL unread as read (ignores notification_ids).

sb notifications show

Show the full body of a single notification.

Arguments:

  • notification_id (required)

Flags:

  • --json — Emit JSON

sb notion

Notion — bidirectional workspace sync, webhooks, and writeback.

sb notion push

Append local markdown content to a Notion page (write_enabled required).

Arguments:

  • page_id (required)

Flags:

  • --file, -f — Markdown file to push (reads stdin if omitted)
  • --dry-run

sb notion status

Show sync state for all tracked Notion pages.

Flags:

  • --json

sb notion sync

Pull Notion pages into the vault (incremental or selective).

Flags:

  • --mode — Sync mode: pull | round-trip
  • --page-id — Specific page IDs for round-trip mode
  • --json
  • --write-vault, --no-write-vault
  • --dry-run

sb notion webhook-serve

Start the Notion webhook receiver server.

Flags:

  • --host
  • --port
  • --secret

sb obsidian

Obsidian — sync and ingest your vault.

sb obsidian ingest

Ingest your Obsidian vault into SecondBrain memory.

Flags:

  • --incremental — Only ingest notes modified in the last 24h
  • --hours — Hours back for incremental ingest

sb obsidian status

Show Obsidian vault status (note count, tags, last note).

Flags:

  • --json

sb obsidian watch

Watch Obsidian vault and ingest modified notes periodically (blocking).

Flags:

  • --interval, -i — Poll interval in seconds
  • --hours — Hours back for each poll

sb ollama

Ollama provider and embedding backend diagnostics.

sb ollama status

Flags:

  • --json — Output JSON status

sb onboard

Guided setup for config, workspace scaffolding, promoted knowledge, and first-run readiness.

sb ontology

Ontology and glossary operations.

sb ontology context

Print the schema prompt context for LLM injection.

sb ontology glossary-add

Add or update a single glossary term.

Arguments:

  • canonical_id (required)
  • term (required)

Flags:

  • --definition
  • --aliases — Comma-separated aliases

sb ontology glossary-list

List all glossary terms.

Search glossary by term or alias.

Arguments:

  • term (required)

Flags:

  • --limit

sb ontology graph

Print relation graph, optionally filtered to a single entity type.

Flags:

  • --entity — Filter to a specific entity type

sb ontology list

List all registered entity types and relations.

sb ontology seed

Seed all entity types, relations, and glossary terms from canonical definitions.

sb ontology show

Show details for a single entity type including its relations.

Arguments:

  • entity_type (required)

sb ontology stats

Print entity_types, relation_types, and glossary_terms counts as JSON.

sb open-loops

Scan vault for TODO/OPENLOOP markers and output a prioritized list.

sb open-loops-fix

Resolve TODO/OPENLOOP markers by replacing them with DONE.

sb ops

Operator-oriented snapshots and summaries.

sb ops snapshot

Materialize the current operator-state snapshot.

Flags:

  • --json — Output JSON

sb ops work-graph

Show the current ranked work graph.

Flags:

  • --limit — Max ranked items
  • --sync — Also sync into graph tables
  • --json — Output JSON

sb outcome

Outcome Factory — turn context into compounding outcomes.

sb outcome accept

Commit to an outcome (won't be reset by future mining).

Arguments:

  • outcome_id (required)

sb outcome brief

Today's ranked outcome board — evidence, ready artifacts, next steps.

Flags:

  • --top, -n — How many cards to show
  • --with-comms, --no-comms — Read live Outlook mail + calendar
  • --with-operator, --no-operator — Also fold in vault commitments/decisions
  • --write-vault — Persist to vault briefs
  • --json — Emit cards as JSON
  • --vault, -v — Vault path

sb outcome curate

Turn one outcome into a reasoned move: fetch thread + draft + verify.

Arguments:

  • outcome_id (required)

Flags:

  • --no-thread — Skip live thread fetch (mail outcomes only)
  • --json — Emit the curated response as JSON
  • --vault, -v — Vault path

sb outcome draft

Draft the ready artifact for an outcome (LLM, grounded in its evidence).

Arguments:

  • outcome_id (required)

Flags:

  • --write-vault, --no-write-vault — Save the draft into the vault
  • --vault, -v — Vault path

sb outcome list

List outcomes already in the ledger.

Flags:

  • --status, -s — Filter by status
  • --limit, -l
  • --json

sb outcome mine

Scan all signals and show the ranked opportunity table (persists to ledger).

Flags:

  • --with-comms, --no-comms
  • --with-operator, --no-operator
  • --json — Emit raw opportunities as JSON
  • --vault, -v — Vault path

sb outcome park

Park an outcome — keep it, but off the live board.

Arguments:

  • outcome_id (required)

sb outcome reflect

Reflect on what shipped vs got rejected — and (with --apply) retune the scorer.

Flags:

  • --days, -d — Lookback window in days
  • --apply — Persist learned category weights so future mining uses them
  • --write-vault — Save the reflection to vault/00_inbox/briefs/
  • --json — Emit raw report as JSON

sb outcome reject

Discard an outcome with a reason (the factory learns from this).

Arguments:

  • outcome_id (required)

Flags:

  • --note — Why it's not worth doing

sb outcome ship

Mark an outcome shipped (after you've actually done it).

Arguments:

  • outcome_id (required)

Flags:

  • --note — What was actually shipped

sb outcome show

Show one outcome in full — evidence, ready artifact, next step.

Arguments:

  • outcome_id (required)

Flags:

  • --json

sb pack

Generate a versioned Context Pack for the given intent.

sb parity

Parity audit — check capability coverage.

sb parity audit

Show capability coverage table.

Flags:

  • --json — Output as JSON

sb parity list

List capability entries with optional filters.

Flags:

  • --status — Filter by status (implemented/planned/stub/deprecated)
  • --category — Filter by category
  • --json — Output as JSON

sb permissions

Show effective policy mode/rules and optionally evaluate one capability.

sb plan

Shortcut for sb tasks plan.

sb plugins

Plugin manifest and capability operations.

sb plugins disable

Disable one plugin for local runtime registration.

Arguments:

  • plugin_id (required)

Flags:

  • --json

sb plugins enable

Enable one plugin for local runtime registration.

Arguments:

  • plugin_id (required)

Flags:

  • --json

sb plugins list

List built-in and installed plugin manifests.

Flags:

  • --category — Filter by category.
  • --json

sb plugins show

Show one plugin manifest.

Arguments:

  • plugin_id (required)

Flags:

  • --json

sb plugins status

Show plugin enablement status.

Arguments:

  • plugin_id

Flags:

  • --json

sb plugins tools

List plugin-contributed kernel tool capabilities.

Arguments:

  • plugin_id

Flags:

  • --json

sb plugins validate

Validate built-in plugin manifests for duplicate ids and unsafe metadata drift.

Arguments:

  • plugin_id

Flags:

  • --json

sb policy

Policy simulation, task-graph policy preview, and governance diagnostics.

sb policy check

Evaluate tool against the loaded ExecPolicy.

Arguments:

  • tool (required)

Flags:

  • --arg — Tool argument as key=value. Repeat for multiple. Use --arg command='git push --force' for shell-style tools.
  • --json — Emit JSON instead of a table.

sb policy check-net

Evaluate protocol://host against the loaded ExecPolicy network rules.

Arguments:

  • host (required)

Flags:

  • --protocol — http | https | *
  • --json — Emit JSON instead of a table.

sb policy explain

Pretty-print the full explain output for tool.

Arguments:

  • tool (required)

Flags:

  • --arg — Tool argument as key=value. Repeat for multiple.

sb policy list

List every loaded ExecPolicy rule, grouped by decision.

Flags:

  • --json — Emit JSON instead of tables.

sb policy preflight

Run the execution-isolation preflight used by command runtimes.

Arguments:

  • command (required)

Flags:

  • --cwd — Working directory for command containment checks.
  • --root — Allowed root. Defaults to shell_exec.cwd_root.
  • --writable-root — Writable root allowed for write-like commands. Repeatable.
  • --backend — Isolation backend for policy evaluation. Defaults to shell_exec.isolation_backend.
  • --network — Network mode for policy evaluation. Defaults to shell_exec.network_mode.
  • --approved — Simulate human approval for caution-level commands.
  • --allow-unsafe-local — Allow write-like commands under local backend.
  • --json — Output JSON

sb policy preview

Preview task-graph policy bundles (sb policy <task_graph_id> also works).

Arguments:

  • task_graph_id (required)

Flags:

  • --json — Output JSON

sb policy simulate

Dry-run policy classification for commands, tools, or action classes.

Flags:

  • --command — Shell command to classify
  • --tool — Tool name to classify
  • --tool-args — Tool arguments as JSON
  • --action-class — Explicit action class
  • --approval-mode — read_only | safe | full
  • --no-persist — Do not store the simulation
  • --json — Output JSON

sb policy validate

Validate a .rules file. Exits 0 on success, 2 on failure.

Arguments:

  • path

sb privacy

Privacy and trust audit commands.

sb privacy audit

Audit privacy posture, credentials, connectors, approvals, and permissions.

Flags:

  • --permission-mode — Override effective permission mode: ask|plan|auto_edit|bypass.
  • --approval-mode — Approval mode alias: read_only|safe|full.
  • --stale-approval-hours — Age threshold for stale pending approval requests.
  • --fail-on-warn — Exit non-zero when warnings or risks are present.
  • --fail-on-risk — Exit non-zero when risk findings are present.
  • --json — Emit machine-readable JSON.

sb proactivity

Proactivity level — controls how autonomously SecondBrain acts.

sb proactivity get

Show the current proactivity level.

sb proactivity list

Show all available proactivity levels.

sb proactivity review

Show proactive loops and nudges derived from operator state.

Flags:

  • --json — Output JSON

sb proactivity set

Set the proactivity level (persisted to config.yaml).

Arguments:

  • level (required)

sb profile

User profile — view and edit vault/USER.md.

sb profile edit

Open vault/USER.md in your default editor.

sb profile show

Display your current user profile parsed from vault/USER.md.

sb project

Project mode — long-horizon coding-agent scaffolding (.sb/ + features.json + progress.md).

sb project enhance

Add missing non-destructive project-mode scaffold files.

Flags:

  • --root — Project root (default: cwd)
  • --dry-run — Preview files without writing
  • --json — JSON output

sb project finalize

End-of-session gate: smoke → commit on green, stash on red, log progress.

Arguments:

  • summary (required)

Flags:

  • --root — Project root (default: cwd)
  • --smoke-timeout — Seconds to wait for .sb/smoke.sh
  • --trace-id — Trace id to include in the progress evidence
  • --session-id — Session id to include in the progress evidence
  • --quality-run-id — Quality run id to include in the progress evidence
  • --replay-case-id — Replay case id to include in the progress evidence
  • --verify — Verification command to include in the progress evidence
  • --json — JSON output

sb project init

Bootstrap project-mode scaffolding (.sb/ + init.sh + features.json + progress.md).

Arguments:

  • name

Flags:

  • --root — Project root (default: cwd)
  • --no-git — Skip the initial git commit
  • --dry-run — Preview files without writing
  • --json — JSON output

sb project status

Show the startup brief: progress tail, recent commits, next feature.

Flags:

  • --root — Project root (default: cwd)
  • --json — JSON output

sb project upgrade

Apply additive project-mode scaffold upgrades.

Flags:

  • --root — Project root (default: cwd)
  • --dry-run — Preview files without writing
  • --json — JSON output

sb projects

Project management operations.

sb projects brief

Show where a project stands based on linked commitments, decisions, and meetings.

Arguments:

  • project_id (required)

Flags:

  • --task-limit — Max commitments/tasks to inspect
  • --decision-limit — Max decisions to inspect
  • --meeting-limit — Max meeting records to inspect
  • --json — Output as JSON

sb projects context

Compile project instructions and scoped knowledge for chat context.

Arguments:

  • project_id (required)

Flags:

  • --max-sources — Max knowledge sources
  • --json — Output as JSON

sb projects context-audit

Audit whether a project has enough layered context for reliable chat.

Arguments:

  • project_id (required)

Flags:

  • --knowledge-source-limit — Warn when active project sources exceed this count
  • --json — Output as JSON

sb projects create

Create a new project.

Arguments:

  • name (required)

Flags:

  • --description, -d — Project description
  • --priority, -p — Priority: low|medium|high|critical
  • --deadline — Deadline (YYYY-MM-DD)
  • --instructions — Project-specific response and operating instructions
  • --visibility — Workspace visibility: private|shared|organization
  • --knowledge-mode — Knowledge mode: curated|retrieval
  • --goal — Project goal (repeatable)

sb projects knowledge

Project knowledge-base operations.

sb projects knowledge add

Add or update a scoped knowledge source for a project.

Arguments:

  • project_id (required)
  • source_ref (required)

Flags:

  • --type — Source type: file|url|text|wiki|memory|context_pack
  • --title — Display title
  • --summary — Short source summary
  • --tag — Source tag (repeatable)
  • --metadata-json — Extra metadata JSON object
  • --json — Output as JSON

sb projects knowledge list

List scoped knowledge sources for a project.

Arguments:

  • project_id (required)

Flags:

  • --include-inactive — Include inactive sources
  • --json — Output as JSON

sb projects list

List all projects.

Flags:

  • --status, -s — Filter by status

sb projects members

Project access and collaboration metadata.

sb projects members add

Add or update a project member.

Arguments:

  • project_id (required)
  • member_id (required)

Flags:

  • --role — Role: can_use|can_edit|owner
  • --name — Display name
  • --json — Output as JSON

sb projects members list

List project members and access roles.

Arguments:

  • project_id (required)

Flags:

  • --include-inactive — Include inactive members
  • --json — Output as JSON

sb projects sessions

Project chat-session links.

Link an existing session to a project workspace.

Arguments:

  • project_id (required)
  • session_id (required)

Flags:

  • --title — Session title
  • --role — Session role
  • --metadata-json — Extra metadata JSON object
  • --json — Output as JSON

sb projects sessions list

List chat sessions linked to a project workspace.

Arguments:

  • project_id (required)

Flags:

  • --limit — Max sessions
  • --json — Output as JSON

sb projects status

Show project status.

Arguments:

  • project_id

Flags:

  • --json — Output as JSON

sb projects update

Update a project.

Arguments:

  • project_id (required)

Flags:

  • --status, -s — New status
  • --priority, -p — New priority
  • --description, -d — New description
  • --deadline — New deadline
  • --instructions — Replace project instructions
  • --visibility — Workspace visibility: private|shared|organization
  • --knowledge-mode — Knowledge mode: curated|retrieval
  • --add-milestone — Add milestone title
  • --milestone-due — Milestone due date

sb prompt

Preview the generated system prompt for debugging.

sb prompts

Prompt management, tagging, and replay.

sb prompts diff

Show a unified diff between two versions of a prompt.

Arguments:

  • prompt_id (required)
  • v1 (required)
  • v2 (required)

sb prompts find-legacy

Find likely inline prompt anti-patterns in Python files.

Flags:

  • --root — Root directory to scan

sb prompts fingerprint

Show fingerprint metadata for a prompt.

Arguments:

  • prompt_id (required)

Flags:

  • --json — Output JSON

sb prompts grep

Search prompt bodies for text.

Arguments:

  • text (required)

sb prompts list

List known prompts with version, source, tags, and engine.

Flags:

  • --tag, -t — Filter by tag
  • --kind — Filter by kind
  • --status — Filter by status
  • --engine — Filter by engine

sb prompts promote

Assert a prompt carries a required tag.

Arguments:

  • prompt_id (required)

Flags:

  • --tag, -t — Required tag for promotion

sb prompts registry-list

List prompts in PromptRegistryV2.

Flags:

  • --json — Output JSON

sb prompts render

Render a prompt with variables.

Arguments:

  • prompt_id (required)

Flags:

  • --vars — JSON variables, e.g. {"k":"v"}
  • --version — Prompt version

sb prompts replay

Replay a prompt against the configured provider.

Arguments:

  • prompt_id (required)

Flags:

  • --input, -i — Input text to replay
  • --vars — JSON variables for prompt rendering
  • --tag, -t — Require this tag before replay

sb prompts show

Print the prompt spec content.

Arguments:

  • prompt_id (required)

Flags:

  • --version — Prompt version

sb prompts validate

Validate bundled and local prompt specs.

Arguments:

  • prompt_id

sb prompts versions

List known versions for a prompt.

Arguments:

  • prompt_id (required)

sb providers

LLM provider health and diagnostics.

sb providers health

Show provider health stats: call counts, errors, and latency.

Flags:

  • --since — Look back N hours
  • --json — Output JSON

sb providers local-profiles

Rank local model profiles against detected hardware and task needs.

Flags:

  • --context — Minimum context tokens required for the task.
  • --requires-tools — Require native tool/function calling support.
  • --allow-cpu, --no-cpu — Allow CPU fallback when no suitable accelerator memory is detected.
  • --profile — Prefer a specific local model profile id when it is available.
  • --json — Output JSON

sb providers matrix

Show provider capabilities and setup requirements.

Flags:

  • --json — Output JSON

sb providers routes

Show provider routing, fallback, failure, and cost observations.

Flags:

  • --since — Look back N hours
  • --limit — Events to return
  • --json — Output JSON

sb quality

Unified quality control plane for runs, suites, and gates.

sb quality gate

Show the latest composite gate decision.

Flags:

  • --surface — Relevant surface to evaluate
  • --json — Output JSON

sb quality harness

Offline operating-harness audits, reviews, and GC proposals.

sb quality harness audit

Audit the local operating harness for agent-readiness gaps.

Flags:

  • --max-file-lines — Line threshold for agent-hostile files.
  • --changed-only — Only scan files currently changed in git status.
  • --json — Output JSON

sb quality harness enrich

Enrich persisted harness records with local git context.

Flags:

  • --limit — Maximum records to enrich.
  • --filter — Only include paths containing text.
  • --json — Output JSON

sb quality harness export

Export persisted harness records as JSON or per-finding Markdown.

Flags:

  • --format — Export format: json or md-dir.
  • --output — Directory for --format md-dir output.
  • --json — Output JSON

sb quality harness gc

Propose harness improvements from recent failures and repeated review pressure.

Flags:

  • --last — Lookback label for the GC report.
  • --json — Output JSON

sb quality harness process

Process persisted harness candidates into findings.

Flags:

  • --run-id — Source scan run id label.
  • --changed-only — Only process records for files currently changed in git status.
  • --matchers — Comma-separated matcher ids to process.
  • --limit — Maximum records to process.
  • --filter — Only include paths containing text.
  • --json — Output JSON

sb quality harness revalidate

Revalidate persisted harness findings against current source.

Flags:

  • --force — Recompute existing verdicts.
  • --matchers — Comma-separated matcher ids to revalidate.
  • --limit — Maximum records to revalidate.
  • --filter — Only include paths containing text.
  • --json — Output JSON

sb quality harness review

Run one deterministic harness review role.

Flags:

  • --role — Review role: reliability, security-policy, runtime-contract, frontend-ux, docs-help, test-strategy
  • --changed-only — Only review files currently changed in git status.
  • --max-file-lines — Line threshold for runtime-contract reviews.
  • --json — Output JSON

sb quality harness scan

Scan source files into persisted harness candidates.

Flags:

  • --changed-only — Only scan files currently changed in git status.
  • --matchers — Comma-separated matcher ids to run.
  • --limit — Maximum files to scan.
  • --filter — Only include paths containing text.
  • --json — Output JSON

sb quality harness status

Show persisted harness pipeline status.

Flags:

  • --json — Output JSON

sb quality improvement-issues

List recurring failure clusters that need gated improvement work.

Flags:

  • --status — Filter by issue status
  • --limit — Issues to return
  • --json — Output JSON

sb quality promote-replay

Promote a proposed replay case into an internal replay eval suite.

Arguments:

  • case_id (required)

Flags:

  • --notes — Operator note for the promotion
  • --force — Promote again if already promoted
  • --json — Output JSON

sb quality replay-cases

List proposed and promoted private replay cases.

Flags:

  • --status — Filter by replay case status
  • --limit — Cases to return
  • --json — Output JSON

sb quality replay-create

Create a replay case from a quality run, trace, or session.

Flags:

  • --quality-run-id — Source quality run id
  • --trace-id — Source trace id
  • --session-id — Source session id
  • --notes — Operator note for the replay case
  • --json — Output JSON

sb quality replay-results

List measured replay executions linked into quality evidence.

Flags:

  • --limit — Results to return
  • --json — Output JSON

sb quality run-replay

Execute a replay case and persist measured pass/fail evidence.

Arguments:

  • case_id (required)

Flags:

  • --json — Output JSON

sb quality runs

List quality runs and suite artifacts.

Flags:

  • --limit — Runs to return
  • --json — Output JSON

sb quality show

Show one quality run, trace, or suite artifact.

Arguments:

  • quality_run_id (required)

Flags:

  • --json — Output JSON

sb quality suites

List suite health across deterministic, judged, and scenario evals.

Flags:

  • --json — Output JSON

sb quality summary

Show the aggregated control-plane summary.

Flags:

  • --limit — Runs to aggregate
  • --json — Output JSON

sb questions

Pending questions — deferred Q&A queue for the operator (ask, answer, cancel).

sb questions answer

Attach an answer to a pending question.

Arguments:

  • qid (required)
  • text (required)

sb questions ask

Record a deferred question for the operator.

Arguments:

  • prompt (required)

Flags:

  • --source — Originating module / channel.
  • --channel-id
  • --user-id
  • --ttl — Auto-expire after: s, m, h, d.
  • --metadata — JSON object of extra metadata.

sb questions cancel

Mark a pending question as cancelled.

Arguments:

  • qid (required)

Flags:

  • --reason — Why this was cancelled.

sb questions list

List pending (or status-filtered) questions.

Flags:

  • --status — Filter to one of: answered, cancelled, expired, pending.
  • --include-expired — Include expired entries.
  • --json — Emit JSON.

sb questions purge

Delete answered / cancelled / expired entries past the cutoff.

Flags:

  • --older-than-days — Delete resolved entries older than N days.

sb questions show

Show one question's full record.

Arguments:

  • qid (required)

Flags:

  • --json — Emit JSON.

sb quota

Token, cost, and context quota diagnostics.

sb quota status

Show quota usage and current enforcement decision for a session.

Flags:

  • --session, -s — Session id to inspect.
  • --json — Output JSON.

sb reflect

Reflect: what happened, what was learned, what should update.

sb registry

Layered registry snapshot operations.

sb registry diff

Diff two snapshots in the same scope.

Arguments:

  • scope (required)

Flags:

  • --from — Baseline snapshot name.
  • --to — Target snapshot name.
  • --limit — Max changes in table output.
  • --json — Output JSON.

sb registry list

List persisted registry snapshots.

Flags:

  • --scope — Filter by scope.
  • --limit — Max snapshots to return.
  • --json — Output JSON.

sb registry promote

Promote one snapshot to another snapshot name (default: active).

Arguments:

  • scope (required)
  • from_snapshot (required)

Flags:

  • --to — Target snapshot name.
  • --source — Source label for promoted snapshot.
  • --description — Optional description.
  • --json — Output JSON.

sb registry set

Write/update a registry snapshot from file payload.

Arguments:

  • scope (required)

Flags:

  • --file — JSON/YAML file with top-level object payload.
  • --snapshot — Snapshot name.
  • --source — Source label.
  • --description — Optional description.
  • --merge — Deep-merge into existing snapshot instead of replacing payload.
  • --json — Output JSON.

sb registry show

Show one registry snapshot payload.

Arguments:

  • scope (required)

Flags:

  • --snapshot — Snapshot name.
  • --json — Output JSON.

sb reindex

Rebuild, verify, and repair vector/hash index integrity.

sb reindex repair

Repair only paths that have vector/hash drift.

Flags:

  • --path, -p — Repair this path (defaults to vault)
  • --include, -i — Comma-separated glob filters
  • --all — Repair every drifted path under the target path.
  • --yes, -y — Skip confirmation prompt

sb reindex verify

Verify drift between tracked files and vector/hash state.

Flags:

  • --path, -p — Path to verify (defaults to vault)
  • --include, -i — Comma-separated glob filters
  • --json — Emit machine-readable JSON output.

sb report

Generate stakeholder reports from data files.

sb report analyze

Analyze an external dataset file and generate a stakeholder report.

Arguments:

  • question (required)

Flags:

  • --source, -s — Path to dataset file (CSV/Excel/JSON).
  • --format, -f — Output format: md, html, docx.
  • --output, -o — Write report to this path (stdout if omitted).
  • --provider, -p — LLM provider.
  • --model, -m — Model override.
  • --no-llm — Skip LLM synthesis; build a structural profile report instead.

sb report join

Generate a join-coverage report before merging two datasets.

Arguments:

  • left (required)
  • right (required)

Flags:

  • --on — Join key column name.
  • --title, -t — Report title override.
  • --format, -f — Output format: md, html, docx.
  • --output, -o — Write to this path (stdout if omitted).

sb report profile

Generate a data-quality profile report for a dataset file.

Arguments:

  • file (required)

Flags:

  • --title, -t — Report title override.
  • --format, -f — Output format: md, html, docx.
  • --output, -o — Write to this path (stdout if omitted).

sb research

Research and synthesis operations.

sb research brief

Generate a daily 'What Matters' brief from configured sources.

Flags:

  • --sources, -s
  • --charter, -c
  • --date
  • --daily
  • --json — Output as JSON

sb research compare

Compare two documents.

Arguments:

  • path1 (required)
  • path2 (required)

sb research deep

Run deep internet research on a topic and save a cited report into the vault.

Arguments:

  • topic (required)

Flags:

  • --slug — Filename slug (auto-derived from topic if omitted)
  • --max-queries — Number of web search queries to generate
  • --max-urls — Maximum URLs to fetch in total
  • --per-doc-chars — Max chars to read per source
  • --output-dir — Output dir for the report (default: vault/05_playbooks/research)
  • --save, --no-save — Write the report into the vault (default: on)
  • --json — Print the report as JSON

sb research prototype

Analyze a paper/blog and generate a prototype plan.

Arguments:

  • input_source (required)

Flags:

  • --name, -n
  • --scaffold
  • --json — Output as JSON

sb research summarize

Summarize a document.

Arguments:

  • input_source (required)

Flags:

  • --style — Summary style: bullets|executive|detailed
  • --mode — Summarization mode: auto|llm|stats
  • --json — Output JSON

sb research-brief

Run the harness-governed research brief agent (ContextOS phases 0-8).

sb research-brief contract

Print the IntentContract for support.research.brief as JSON.

sb research-brief manifest

Print the ContextPackManifest the agent would compile for topic.

Arguments:

  • topic (required)

sb research-brief run

Run a research brief through all eight harness phases.

Arguments:

  • topic (required)

Flags:

  • --stage — shadow | canary | partial | full
  • --percentage — Override stage default percentage.
  • --kill-switch — Path to a kill-switch sentinel file (if present, agent is bypassed).
  • --no-verify, --no-no-verify — Skip the critic.verify pre-mark — exercises the side-effect gate.
  • --request-key — Deterministic key for rollout bucket hashing (defaults to topic).
  • --json — Emit a structured JSON envelope.

sb resume

Shortcut for sb tasks resume.

sb rewind

Restore files from a durable checkpoint.

sb route

Route a natural-language query to matching sb commands.

sb routines

Recurring automations (routines): schedules, API triggers, and GitHub webhooks.

sb routines guide

Show how session automations, workflow automations, and webhooks fit together.

sb routines templates

Claude Code Routine templates for SecondBrain.

sb routines templates init

Output a template's prompt ready to paste into the /schedule wizard.

Arguments:

  • name (required)

Flags:

  • --output, -o — Output file path (default: stdout)

sb routines templates list

List all available Routine templates.

Flags:

  • --trigger, -t — Filter by trigger type: schedule | api | github
  • --json

sb routines templates show

Show a template's full prompt and how to deploy it via /schedule.

Arguments:

  • name (required)

Flags:

  • --json

sb run

Run a task graph or an agent pattern.

sb runtime-guard

Governed runtime tool-contract operations.

sb runtime-guard approve-contract

Approve one current live contract as the persisted baseline.

Arguments:

  • surface (required)
  • tool_name (required)

Flags:

  • --approved-by
  • --reason
  • --permission-mode
  • --approval-mode
  • --web
  • --tool-set
  • --skill-root
  • --include-codex-home-skills, --no-include-codex-home-skills
  • --subagents, --no-subagents
  • --json — Output JSON

sb runtime-guard approve-edge

Approve one observed delegation topology edge.

Arguments:

  • edge_id (required)

Flags:

  • --approved-by
  • --reason
  • --json — Output JSON

sb runtime-guard lock

Lock a delegation topology scope so only approved edges are allowed.

Arguments:

  • scope

Flags:

  • --locked-by
  • --reason
  • --json — Output JSON

sb runtime-guard mismatches

Show only new, missing, or mismatched tool contracts.

Flags:

  • --surface
  • --permission-mode
  • --approval-mode
  • --web
  • --tool-set
  • --skill-root
  • --include-codex-home-skills, --no-include-codex-home-skills
  • --subagents, --no-subagents
  • --fail-on-drift, --no-fail-on-drift — Exit 1 when any new, missing, or mismatched contract is found.
  • --json — Output JSON

sb runtime-guard snapshot

Approve the current live contracts as the persisted baseline.

Flags:

  • --surface
  • --approved-by
  • --reason
  • --permission-mode
  • --approval-mode
  • --web
  • --tool-set
  • --skill-root
  • --include-codex-home-skills, --no-include-codex-home-skills
  • --subagents, --no-subagents
  • --json — Output JSON

sb runtime-guard tools

List current tool contracts compared with approved baselines.

Flags:

  • --surface
  • --permission-mode
  • --approval-mode
  • --web
  • --tool-set
  • --skill-root
  • --include-codex-home-skills, --no-include-codex-home-skills
  • --subagents, --no-subagents
  • --json — Output JSON

sb runtime-guard topology

List observed and approved delegation topology edges.

Flags:

  • --scope
  • --status
  • --limit
  • --json — Output JSON

sb runtime-guard unlock

Unlock a delegation topology scope.

Arguments:

  • scope

Flags:

  • --reason
  • --json — Output JSON

sb schema

Schema registry operations.

sb schema deprecate

Arguments:

  • schema_id (required)

Flags:

  • --replacement

sb schema list

List all registered schemas, optionally filtered by namespace.

Flags:

  • --namespace

sb schema register

Arguments:

  • namespace (required)

Flags:

  • --version
  • --schema-json

sb schema show

Arguments:

  • schema_id (required)

sb scorecard

sb sdk

Run tasks via the Claude Agent SDK.

sb sdk configs

List available agent configuration presets.

sb sdk fork

Fork an existing session — explore an alternate branch with full prior context.

Arguments:

  • session_id (required)
  • prompt (required)

Flags:

  • --mode, -m
  • --json

sb sdk run

Run a prompt through the Agent SDK with SecondBrain context.

Arguments:

  • prompt (required)

Flags:

  • --mode, -m — Agent config: readonly | research | coding
  • --resume — Resume a session by ID
  • --fork — Fork the resumed session instead of continuing it
  • --max-turns — Max agent turns
  • --json — Output result as JSON
  • --no-context — Skip CLAUDE.md injection
  • --interactive — Allow the agent to call AskUserQuestion mid-run
  • --with-agents — Register subagents from .claude/agents/*.md and add Agent tool
  • --no-default-hooks — Disable built-in event-log/vault-audit/content-safety hooks

sb security

Security and consent operations.

Arguments:

  • subject_id (required)

Arguments:

  • subject_id (required)
  • scope (required)

Flags:

  • --granted, --denied

sb sentinel

Workspace sentinels — pause / dnd / presenter signals that gate autonomous work.

sb sentinel clear

Clear a sentinel.

Arguments:

  • name (required)

sb sentinel known

List the names recognized by built-in consumers.

sb sentinel list

List live (non-expired) sentinels.

Flags:

  • --json — Emit JSON.

sb sentinel set

Set or refresh a sentinel.

Arguments:

  • name (required)

Flags:

  • --reason — Why this sentinel was set.
  • --duration — TTL: s, m, h, d. Defaults to no expiry.
  • --until — ISO timestamp at which this sentinel expires.

sb sentinel show

Show a single sentinel.

Arguments:

  • name (required)

Flags:

  • --json — Emit JSON.

sb serve

Start localhost API server for UI clients (command parity daemon).

sb serve-token

Manage the local serve bearer token without exposing it by default.

sb serve-token env

Print a shell export line for Docker Compose or one-off clients.

Flags:

  • --create, --no-create — Create and persist a token if none exists.

sb serve-token init

Create the local auth file if needed, preserving a strong existing token.

Flags:

  • --token — Explicit token to persist; must be strong enough for managed use.
  • --print-token — Print the secret token after initializing.
  • --json — Emit JSON output.

sb serve-token rotate

Rotate the persisted local serve token.

Flags:

  • --token — Explicit replacement token; omit to generate a new one.
  • --print-token — Print the new secret token after rotation.
  • --json — Emit JSON output.

sb serve-token show

Print the active secret token for scripts. Use intentionally.

Flags:

  • --create — Create and persist a token if none exists.

sb serve-token status

Inspect serve-token state without revealing the token.

Flags:

  • --json — Emit JSON output.

sb session-log

Durable session log — list/replay/fork items by seq.

sb session-log clear

Delete all items for a session.

Arguments:

  • session_id (required)

Flags:

  • --db — Path to the session SQLite file.
  • --yes, -y — Skip the confirmation prompt.
  • --json

sb session-log exec-events

Replay every event for one execution, in append order.

Arguments:

  • exec_id (required)

Flags:

  • --db — Path to the SQLite file.
  • --json

sb session-log exec-status

Show the latest state for one execution.

Arguments:

  • exec_id (required)

Flags:

  • --db — Path to the SQLite file.
  • --json

sb session-log exec-tree

Render the parent → child execution tree rooted at exec_id.

Arguments:

  • exec_id (required)

Flags:

  • --db — Path to the SQLite file.
  • --max-depth — Maximum tree depth to walk (defensive against malformed parent loops).
  • --json

sb session-log executions

List pending (and optionally completed/failed) executions for a session.

Arguments:

  • session_id (required)

Flags:

  • --db — Path to the SQLite file.
  • --pending-only — Only show executions still in PENDING state (i.e. resumable).
  • --json

sb session-log fork

Fork a session at the given checkpoint into a new session id.

Arguments:

  • src (required)
  • dest (required)

Flags:

  • --src-seq — Seq to fork at (inclusive). 0 means fork the whole history.
  • --db — Path to the session SQLite file.
  • --json

sb session-log items

List all items for a session in seq order.

Arguments:

  • session_id (required)

Flags:

  • --db — Path to the session SQLite file.
  • --json — Emit JSON instead of human output.

sb session-log pending-sessions

List sessions that have any PENDING execs.

Flags:

  • --db — Path to the SQLite file.
  • --show-execs — Also list each pending exec_id under its session.
  • --json

sb session-log replay

Replay items strictly after the given seq.

Arguments:

  • session_id (required)

Flags:

  • --last-seq — Last seq the client saw; the log replays items strictly after this seq.
  • --db — Path to the session SQLite file.
  • --json

sb session-log resume-check

Check whether agent_id can legitimately resume exec_id.

Arguments:

  • exec_id (required)
  • agent_id (required)

Flags:

  • --db — Path to the SQLite file.
  • --json

sb session-log tail

Show the last item and whether the session is paused on a confirmation.

Arguments:

  • session_id (required)

Flags:

  • --db — Path to the session SQLite file.
  • --json

sb sessions

Chat, durable background session, and recurring scout automation operations.

sb sessions approve

Approve a pending background-session request.

Arguments:

  • session_id (required)
  • request_id (required)

Flags:

  • --scope — agent | person | global
  • --persist-rule, --no-persist-rule
  • --json — Output JSON

sb sessions attach

Follow session events until the session reaches a terminal state.

Arguments:

  • session_id (required)

Flags:

  • --interval-s — Polling interval
  • --limit — Max events to print before exiting
  • --json — Output JSON

sb sessions automations

Recurring background session automations.

sb sessions automations clear-invoke-token

Remove per-automation invoke token.

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb sessions automations create

Create a recurring background-session automation.

Arguments:

  • name (required)

Flags:

  • --prompt — Recurring scout/agent prompt
  • --schedule — Schedule: interval: or cron:<5-field>
  • --engine — native | claude | codex
  • --runner-kind — chat | agentic | bridge
  • --scout-kind — Short automation role label
  • --sleep-time — Configure the automation to run the governed sleep-time lane.
  • --cwd — Working directory
  • --provider — Provider for native automations
  • --model — Optional model override
  • --web-mode — on | off | auto
  • --top-k — Vault chunks for context
  • --profile — Native chat profile
  • --permission-mode — Permission mode
  • --approval-mode — Bridge approval mode
  • --think-level — Thinking level
  • --context-topic — Optional vault context topic
  • --no-context — Skip vault context injection
  • --task-title — Optional scheduled task title
  • --task-description — Optional scheduled task description
  • --task-priority — low | medium | high | critical
  • --project-id — Optional project id for seeded task
  • --workflow-id — Optional workflow target
  • --workflow-inputs-json — Optional workflow inputs JSON object
  • --paused — Create automation in paused state
  • --generate-invoke-token — Issue per-automation token for POST .../run (shown once).
  • --json — Output JSON

sb sessions automations delete

Delete one recurring background-session automation definition.

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb sessions automations list

List recurring background-session automations.

Flags:

  • --status — Optional automation status filter
  • --limit — Max automations to return
  • --json — Output JSON

sb sessions automations pause

Pause one recurring background-session automation.

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb sessions automations resume

Resume one recurring background-session automation.

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb sessions automations rotate-invoke-token

Rotate the per-automation invoke token for API triggers.

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb sessions automations run

Run one recurring automation now or launch all currently due scout automations.

Arguments:

  • automation_id

Flags:

  • --due-only — Run all due automations.
  • --dry-run — Preview without launching a session
  • --detach, --no-detach — Launch in background
  • --as-of — Optional override timestamp
  • --limit — Maximum due automations to run
  • --json — Output JSON

sb sessions automations show

Show one recurring background-session automation.

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb sessions branch

Create a branched background session from a durable checkpoint.

Arguments:

  • session_id (required)

Flags:

  • --from-checkpoint — Checkpoint id to resume from
  • --goal — Goal for the branched background session
  • --detach, --no-detach — Run in the background
  • --json — Output JSON

sb sessions cancel

Cancel a background session.

Arguments:

  • session_id (required)

Flags:

  • --json — Output JSON

sb sessions list

List recent chat sessions with optional background-session state.

Flags:

  • --limit — Max sessions to return
  • --status — Optional background session status filter
  • --json — Output JSON

sb sessions pause

Pause a background session without marking it terminal.

Arguments:

  • session_id (required)

Flags:

  • --json — Output JSON

sb sessions recover

Apply an operator recovery action to a durable background session.

Arguments:

  • session_id (required)

Flags:

  • --action — Recovery action: resume | expire | cancel
  • --detach, --no-detach — Resume in the background
  • --json — Output JSON

sb sessions resume

Resume a queued, recovering, or approved-awaiting-approval session.

Arguments:

  • session_id (required)

Flags:

  • --detach, --no-detach — Resume in the background
  • --json — Output JSON

sb sessions show

Show one session with messages, events, tool calls, and background metadata.

Arguments:

  • session_id (required)

Flags:

  • --tool-calls — Include recent tool-call rows (trace-aware) in output.
  • --tool-calls-limit — Max recent tool-call rows when --tool-calls is set.
  • --verbose, -v — Show persisted diagnostic event details.
  • --json — Output JSON

sb sessions start

Start a durable background session.

Arguments:

  • goal (required)

Flags:

  • --engine — native | claude | codex
  • --profile — Agent profile for native chat sessions
  • --provider — Provider for native sessions
  • --model — Optional model override
  • --cwd — Working directory
  • --web-mode — on | off | auto
  • --top-k — Vault chunks for context
  • --permission-mode — Permission mode
  • --approval-mode — Bridge approval mode
  • --think-level — Thinking level
  • --sleep-time — Run the governed sleep-time lane instead of a normal background chat.
  • --session-id — Optional durable session id
  • --detach, --no-detach — Run in the background
  • --max-retries — Retry budget
  • --expires-at — Optional ISO-8601 expiry
  • --json — Output JSON

sb sessions sweep

Mark orphaned background sessions as failed.

Flags:

  • --queued-timeout — Kill sessions stuck in 'queued' longer than this many seconds (default 5 min).
  • --heartbeat-timeout — Kill 'running' sessions whose heartbeat is older than this many seconds AND whose process is unreachable (default 2.5 h).
  • --dry-run, -n — Report what would be swept without mutating anything.
  • --notify — Emit 'session.swept' notifications for each reaped row.
  • --json — Emit JSON.

sb simulate

Synthetic simulation suites and offline conversation replay.

sb simulate conversations

Build and run synthetic conversation suites.

sb simulate conversations build

Validate and materialize one simulation suite.

Arguments:

  • suite (required)

Flags:

  • --pack — Pack: smoke|core|regression
  • --out — Output directory override
  • --json — Output JSON

sb simulate conversations run

Run one simulation suite across the selected surfaces.

Arguments:

  • suite (required)

Flags:

  • --surface — Execution surface: chat|autotune|both
  • --pack — Pack: smoke|core|regression
  • --out — Output directory override
  • --json — Output JSON

sb skills

Local skill-pack discovery operations.

sb skills add

Import a skill folder or zip archive into the skill root.

Arguments:

  • source (required)

Flags:

  • --path — Target skill root directory (default: ./skills)
  • --json — Output JSON

sb skills apply-edits

Apply bounded SkillOpt-style edits to a skill.

Arguments:

  • skill_name (required)

Flags:

  • --edits — JSON or YAML file containing SkillOpt-style edits
  • --path — Skill root directory (repeatable)
  • --edit-budget — Maximum number of edits to apply
  • --candidate-output — Write candidate SKILL.md content to this file
  • --write, --dry-run — Write the candidate back to SKILL.md. Defaults to dry-run.
  • --json — Output JSON

sb skills build-candidate

Materialize a candidate SKILL.md from a bounded edit proposal.

Arguments:

  • skill_name (required)

Flags:

  • --edits — JSON or YAML SkillOpt-style edit proposal
  • --path — Skill root directory (repeatable)
  • --run-state — Update an optimizer run directory or run.json file with this candidate
  • --candidate-output — Candidate SKILL.md path to write. Defaults to run candidates directory.
  • --apply-report-output — Optional JSON apply report path. Defaults next to the candidate.
  • --edit-budget — Maximum number of edits to apply
  • --json — Output JSON

sb skills create

Scaffold a new optimizer-ready skill folder.

Arguments:

  • name (required)

Flags:

  • --path — Target skill root directory (default: ./skills)
  • --description, -d — Skill description
  • --trigger — Concrete activation phrase (repeatable)
  • --negative-trigger — Phrase that should not activate this skill (repeatable)
  • --instruction — Workflow instruction to include in SKILL.md (repeatable)
  • --example — Example request or behavior to include in SKILL.md (repeatable)
  • --success-criterion — Observable success criterion for this skill (repeatable)
  • --failure-mode — Known failure mode or boundary condition (repeatable)
  • --validation-command — Validation command for this skill (repeatable)
  • --allowed-tool — Tool or tool namespace this skill may use (repeatable)
  • --required-tool — Tool or tool namespace required for this skill (repeatable)
  • --trust-level — Skill trust level: local, reviewed, trusted, experimental, or deprecated
  • --safety-class — Skill safety class: read_only, standard, write, network, privileged, or destructive
  • --eval-cases, --no-eval-cases — Create references/eval_cases.yaml with train/selection/test cases
  • --slow-update, --no-slow-update — Include optimizer protected slow-update guidance markers
  • --creation-brief, --no-creation-brief — Create references/creation_brief.yaml with builder intent and capability contract
  • --json — Output JSON

sb skills docs-write

Generate docs/reference/skills.md from SKILL.md frontmatter.

Flags:

  • --path — Skill root directory (repeatable). Defaults to configured skill roots
  • --output — Markdown file to write
  • --include-codex-home, --no-include-codex-home — Also scan $CODEX_HOME/skills if configured
  • --json — Output JSON

sb skills gate-candidate

Gate a candidate skill by strict held-out score improvement.

Arguments:

  • skill_name (required)

Flags:

  • --baseline-score — Held-out selection score for the current baseline skill
  • --candidate-score — Held-out selection score for the candidate skill
  • --path — Skill root directory (repeatable)
  • --candidate — Optional candidate SKILL.md path for provenance
  • --edit-report — Optional apply-edits report to attach to rejection buffer
  • --rejected-buffer — Write rejected-edit buffer entry here when the gate rejects
  • --antahkarana-db — Record the gate outcome into an Antahkarana SQLite database
  • --run-state — Update an optimizer run directory or run.json file with this gate
  • --output — Optional JSON or YAML gate result file to write
  • --min-delta — Minimum positive improvement required above baseline
  • --fail-on-reject, --no-fail-on-reject — Exit non-zero when the candidate is rejected
  • --json — Output JSON

sb skills install

Install a skill pack from PyPI and discover its registered skills.

Arguments:

  • package (required)

Flags:

  • --upgrade, -U — Pass --upgrade to pip
  • --json — Output JSON

sb skills installed

List skill packs installed via PyPI entry points (brain.skills group).

Flags:

  • --json — Output JSON

sb skills list

Discover local skill packs from skill roots.

Flags:

  • --path — Skill root directory (repeatable). Defaults to ./skills
  • --include-codex-home, --no-include-codex-home — Also scan $CODEX_HOME/skills if set
  • --json — Output JSON

sb skills optimize-plan

Draft a validation-gated skill optimization plan.

Arguments:

  • skill_name (required)

Flags:

  • --path — Skill root directory (repeatable)
  • --objective — Optimization objective for this skill
  • --evidence — Scored rollout, replay, or benchmark evidence file (repeatable)
  • --output — Optional JSON or YAML plan file to write
  • --json — Output JSON

sb skills optimize-start

Create a durable skill optimization run directory.

Arguments:

  • skill_name (required)

Flags:

  • --path — Skill root directory (repeatable)
  • --objective — Optimization objective for this skill
  • --evidence — Scored rollout, replay, or benchmark evidence file (repeatable)
  • --run-dir — Directory for optimizer run artifacts
  • --json — Output JSON

sb skills optimize-status

Show durable skill optimization run status.

Arguments:

  • run_state (required)

Flags:

  • --json — Output JSON

sb skills pack

Pack a skill folder into a zip archive for sharing.

Arguments:

  • skill_name (required)

Flags:

  • --path — Skill root directory (repeatable)
  • -o, --output — Output zip path (default: .zip)

sb skills propose-edits

Propose bounded SkillOpt-style edits from scored evidence.

Arguments:

  • skill_name (required)

Flags:

  • --evidence — Scored rollout, replay, or benchmark evidence file (repeatable)
  • --path — Skill root directory (repeatable)
  • --edit-budget — Maximum number of ranked evidence patterns to include
  • --failure-threshold — Normalized score below this is treated as failure evidence
  • --success-threshold — Normalized score at or above this is treated as success evidence
  • --output — Optional JSON or YAML edit proposal file to write
  • --run-state — Update an optimizer run directory or run.json file with this proposal
  • --json — Output JSON

sb skills registry-export

Export the current skill registry payload.

Flags:

  • --json — Output JSON

sb skills registry-refresh

Refresh skill discovery and persist the current manifest set.

Flags:

  • --path — Skill root directory (repeatable). Defaults to configured skill roots
  • --include-codex-home, --no-include-codex-home
  • --json — Output JSON

sb skills run

Prepare a skill-scoped agent prompt package for a task.

Arguments:

  • skill_name (required)
  • task (required)

Flags:

  • --path — Skill root directory (repeatable)
  • --json — Output JSON

sb skills show

Show details for a specific skill.

Arguments:

  • skill_name (required)

Flags:

  • --path — Skill root directory (repeatable)
  • --meta-only — Show frontmatter only
  • --full — Show full SKILL.md content
  • --json — Output JSON

sb skills validate

Validate skill folders against SKILL.md spec rules.

Flags:

  • --path — Skill root directory (repeatable). Defaults to ./skills
  • --strict — Enable strict validation rules
  • --json — Output JSON

sb slack

Slack — send messages and run a Socket Mode bot.

sb slack auth

Save your Slack bot token to keys.env.

sb slack channels

List Slack channels the bot is in.

sb slack listen

Start a Slack Socket Mode bot (blocking).

Flags:

  • --briefing — Enable /briefing command in Slack

sb slack send

Send a message to a Slack channel or DM.

Arguments:

  • channel (required)
  • message (required)

Flags:

  • --force — Send even if proactivity level < assistant

sb snapshot

Export a standard brain-state snapshot to JSON.

sb social

Review-first social publishing operations.

sb social x

X content draft and review operations.

sb social x drafts

Local X draft queue; no external publishing.

sb social x drafts create

Create a local X draft record for operator review.

Flags:

  • --text — Draft text to store.
  • --text-file — Read draft text from a local file.
  • --pillar — Content pillar for the draft.
  • --format — short_post | mini_thread | reply_template | agent_engineering_note
  • --source-ref — Evidence reference; pass multiple times for multiple sources.
  • --audience — Target audience label.
  • --cta-posture — CTA posture.
  • --link-policy — Link posture.
  • --confidence — Evidence confidence.
  • --quality-score — Optional draft quality score from 0 to 100.
  • --json — Output JSON

sb social x drafts history

Show governed X publish action history for one social draft.

Arguments:

  • draft_id (required)

Flags:

  • --json — Output JSON

sb social x drafts list

List local X drafts by review status.

Flags:

  • --status, -s — proposed | approved | needs_revision | rejected | published | archived | all
  • --json — Output JSON

sb social x drafts queue-publish

Queue an approved X draft as a governed publish action.

Arguments:

  • draft_id (required)

Flags:

  • --json — Output JSON

sb social x drafts review

Record an operator review decision for one local X draft.

Arguments:

  • draft_id (required)

Flags:

  • --decision — approve | request_changes | reject | archive
  • --reviewer — Reviewer name or id.
  • --notes — Review notes.
  • --json — Output JSON

sb social x drafts show

Show one local X draft.

Arguments:

  • draft_id (required)

Flags:

  • --json — Output JSON

sb stack

Optional integration stack health checks.

sb stack check

Flags:

  • --json — Output JSON status

sb status

Shortcut for sb tasks status.

sb superpowers

Superpowers — manage durable Claude/Codex bridge sessions and related state.

sb superpowers sessions

Manage durable Claude/Codex bridge sessions.

sb superpowers sessions clear

Remove all stored sessions (use with care).

Flags:

  • --bridge, -b — Clear only sessions for this bridge
  • --yes, -y — Skip confirmation prompt

sb superpowers sessions list

List persisted bridge sessions.

Flags:

  • --bridge, -b — Filter to one bridge: claude | codex
  • --json — Emit JSON

sb superpowers sessions reset

Forget a bridge session (the next run will start a fresh conversation).

Arguments:

  • bridge (required)
  • name (required)

sb superpowers sessions show

Show details for a single bridge session.

Arguments:

  • bridge (required)
  • name (required)

Flags:

  • --json — Emit JSON

sb task

Singular task workstream shortcuts.

sb task start

Shortcut for sb tasks start.

Arguments:

  • goal (required)

Flags:

  • --worktree, --no-worktree
  • --permission-mode
  • --sandbox-mode
  • --json — Output JSON

sb tasks

Task planning and execution.

sb tasks act

Alias for sb tasks run using the plan/act vocabulary.

Arguments:

  • task_graph_id (required)

Flags:

  • --yes, -y — Allow write actions
  • --approve-plan
  • --compiled-controls, --no-compiled-controls
  • --permission-mode
  • --enforce-template-gates, --no-enforce-template-gates
  • --strict — Block if codified-context drift is detected

sb tasks approve

Approve a plan artifact so it can be executed.

Arguments:

  • task_graph_id (required)

sb tasks classify

Classify intent and risk tier for deterministic planning.

Arguments:

  • goal (required)

Flags:

  • --json — Output JSON

sb tasks explain

Explain why each task step is runnable or blocked.

Arguments:

  • task_graph_id (required)

Flags:

  • --yes, -y
  • --compiled-controls, --no-compiled-controls
  • --permission-mode
  • --next-only
  • --enforce-template-gates, --no-enforce-template-gates
  • --json — Output JSON

sb tasks gate-approve

Record a human approval decision for a task gate.

Arguments:

  • task_graph_id (required)
  • gate_id (required)

Flags:

  • --decision — approved|denied|requested
  • --approver — Approver identity
  • --rationale — Decision rationale
  • --expires-minutes

sb tasks gate-list

List approval gate decisions for a task graph/session.

Arguments:

  • task_graph_id (required)

Flags:

  • --json — Output JSON

sb tasks plan

Plan a deterministic task graph.

Arguments:

  • goal (required)

Flags:

  • --delegate-step — Route selected step_id via a2a_delegate capability (repeatable).
  • --a2a-to-agent
  • --a2a-via
  • --a2a-agent-card
  • --a2a-model
  • --a2a-payload-json
  • --strict — Block if codified-context drift is detected
  • --json — Output as JSON

sb tasks policy

Preview active policy bundles selected for this task graph's intent/risk scope.

Arguments:

  • task_graph_id (required)

Flags:

  • --json — Output JSON

sb tasks resume

Resume from pending/blocked/failed steps without rerunning completed steps.

Arguments:

  • task_graph_id (required)

Flags:

  • --yes, -y — Allow write actions
  • --approve-plan
  • --compiled-controls, --no-compiled-controls
  • --permission-mode
  • --from-step
  • --force-step
  • --dry-run
  • --enforce-template-gates, --no-enforce-template-gates
  • --strict — Block if codified-context drift is detected

sb tasks run

Execute a planned task graph via governed action gateway.

Arguments:

  • task_graph_id (required)

Flags:

  • --yes, -y — Allow write actions
  • --approve-plan
  • --compiled-controls, --no-compiled-controls
  • --permission-mode
  • --enforce-template-gates, --no-enforce-template-gates
  • --strict — Block if codified-context drift is detected

sb tasks start

Start a task workstream, optionally preparing an isolated workspace.

Arguments:

  • goal (required)

Flags:

  • --worktree, --no-worktree
  • --permission-mode
  • --sandbox-mode
  • --json — Output JSON

sb tasks status

Show task graph execution status and plan artifact lifecycle state.

Arguments:

  • task_graph_id (required)

Flags:

  • --json — Output JSON

sb tasks templates

List deterministic task templates and risk-tier constraints.

Flags:

  • --json — Output JSON

sb tasks workspace

Task workspace / worktree lifecycle.

sb tasks workspace cleanup

Release and remove the active workspace for a task graph.

Arguments:

  • task_graph_id (required)

Flags:

  • --json — Output JSON

sb tasks workspace list

List task workspaces.

Flags:

  • --active-only
  • --limit
  • --json — Output JSON

sb tasks workspace prepare

Prepare or reuse an isolated workspace for a task graph.

Arguments:

  • task_graph_id (required)

Flags:

  • --permission-mode
  • --sandbox-mode
  • --json — Output JSON

sb tasks workspace promote

Dry-run or apply a safe fast-forward merge from a task worktree.

Arguments:

  • task_graph_id (required)

Flags:

  • --apply — Fast-forward merge the candidate
  • --yes, -y — Required with --apply
  • --require-review — Require --reviewed-by before promotion is considered ready.
  • --reviewed-by — Reviewer identity recorded with a reviewed promotion.
  • --json — Output JSON

sb tasks workspace show

Show the active workspace for a task graph.

Arguments:

  • task_graph_id (required)

Flags:

  • --json — Output JSON

sb tasks workspace status

Report workspace path, git cleanliness, and merge readiness.

Arguments:

  • task_graph_id

Flags:

  • --active-only
  • --limit
  • --json — Output JSON

sb teams

Microsoft Teams — Graph-backed chat reads and sends.

sb teams auth

Store the Microsoft Graph access token in the SecondBrain keychain.

Arguments:

  • value

Flags:

  • --token-env — Environment/keychain name for the Microsoft Graph access token.
  • --json

sb teams auth-doctor

Diagnose Teams Graph authentication setup.

Flags:

  • --json

sb teams bot-manifest

Print a minimal Teams app manifest for the conversational bot adapter.

Flags:

  • --app-id — Microsoft Bot/Entra app ID.
  • --name — Teams app display name.
  • --description — Teams app short description.
  • --json

sb teams bot-status

Show Teams gateway channel adapter configuration status.

Flags:

  • --json

sb teams channel-message

Fetch one Teams channel message by ID.

Arguments:

  • team_id (required)
  • channel_id (required)
  • message_id (required)

Flags:

  • --json

sb teams channel-messages

Read recent messages from one Teams channel.

Arguments:

  • team_id (required)
  • channel_id (required)

Flags:

  • --limit
  • --all-pages — Follow Graph nextLink pages.
  • --max-pages
  • --json

sb teams channel-replies

Read replies under one Teams channel message.

Arguments:

  • team_id (required)
  • channel_id (required)
  • message_id (required)

Flags:

  • --limit
  • --all-pages — Follow Graph nextLink pages.
  • --max-pages
  • --json

sb teams channels

List channels for one Microsoft Teams team.

Arguments:

  • team_id (required)

Flags:

  • --limit
  • --all-pages — Follow Graph nextLink pages.
  • --max-pages
  • --json

sb teams chat-message

Fetch one Teams chat message by ID.

Arguments:

  • chat_id (required)
  • message_id (required)

Flags:

  • --json

sb teams chats

List recent Teams chats.

Flags:

  • --limit
  • --all-pages — Follow Graph nextLink pages.
  • --max-pages
  • --json

sb teams create-planner-task

Create a task in one Microsoft Planner bucket.

Arguments:

  • plan_id (required)
  • bucket_id (required)
  • title (required)

Flags:

  • --due-date-time — ISO 8601 due timestamp.
  • --start-date-time — ISO 8601 start timestamp.
  • --assign — Microsoft Entra user ID to assign. Repeat for multiple assignees.
  • --dry-run — Preview without creating a task.
  • --force — Confirm this external write.
  • --json

sb teams delete-planner-task

Delete a Microsoft Planner task using an ETag-safe write.

Arguments:

  • task_id (required)

Flags:

  • --etag — Planner task ETag. Fetched when omitted.
  • --dry-run — Preview without deleting.
  • --force — Confirm this external write.
  • --json

sb teams device-login

Sign in with Microsoft identity device-code flow and cache Graph tokens.

Flags:

  • --client-id — Microsoft Entra app client ID.
  • --tenant — Tenant, domain, or organizations.
  • --scope — Graph delegated scope. Repeat or use config defaults.
  • --token-env — Keychain token name to store.
  • --json

sb teams joined-teams

List Microsoft Teams teams joined by the signed-in user.

Flags:

  • --limit
  • --all-pages — Follow Graph nextLink pages.
  • --max-pages
  • --json

sb teams login

Use the signed-in Azure CLI account as the Teams token provider.

Flags:

  • --json

sb teams messages

Read recent messages from one Teams chat.

Arguments:

  • chat_id (required)

Flags:

  • --limit
  • --all-pages — Follow Graph nextLink pages.
  • --max-pages
  • --json

sb teams planner-buckets

List Planner buckets for one plan.

Arguments:

  • plan_id (required)

Flags:

  • --json

sb teams planner-plans

List Planner plans owned by one Microsoft 365 group or team.

Arguments:

  • group_id (required)

Flags:

  • --json

sb teams planner-task

Fetch one Microsoft Planner task by ID.

Arguments:

  • task_id (required)

Flags:

  • --json

sb teams planner-task-details

Fetch one Microsoft Planner task details object.

Arguments:

  • task_id (required)

Flags:

  • --json

sb teams planner-tasks

List Planner tasks for one plan.

Arguments:

  • plan_id (required)

Flags:

  • --all-pages — Follow Graph nextLink pages.
  • --max-pages
  • --json

sb teams profile

Read the signed-in Microsoft Graph profile.

Flags:

  • --json

sb teams resolve-channel

Resolve channels inside one Microsoft Teams team.

Arguments:

  • team_id (required)
  • query (required)

Flags:

  • --limit
  • --json

sb teams resolve-chat

Resolve recent chats by topic, type, or ID.

Arguments:

  • query (required)

Flags:

  • --limit
  • --json

sb teams resolve-team

Resolve joined Microsoft Teams teams by visible name or ID.

Arguments:

  • query (required)

Flags:

  • --limit
  • --json

sb teams resolve-users

Resolve Microsoft Entra users for Teams routing and mentions.

Arguments:

  • query (required)

Flags:

  • --limit
  • --json

sb teams send-channel

Send a message to a Teams channel.

Arguments:

  • team_id (required)
  • channel_id (required)
  • message (required)

Flags:

  • --html — Send the message body as HTML.
  • --mention
  • --attachment-json
  • --dry-run — Preview without posting.
  • --force — Confirm this external write.
  • --json

sb teams send-channel-reply

Send a reply under one Teams channel message.

Arguments:

  • team_id (required)
  • channel_id (required)
  • message_id (required)
  • message (required)

Flags:

  • --html — Send the reply body as HTML.
  • --mention
  • --attachment-json
  • --dry-run — Preview without posting.
  • --force — Confirm this external write.
  • --json

sb teams send-chat

Send a message to an existing Teams chat.

Arguments:

  • chat_id (required)
  • message (required)

Flags:

  • --html — Send the message body as HTML.
  • --mention — Mention in USER_ID=Display Name form. Repeat for multiple mentions.
  • --attachment-json — Graph chatMessage attachment JSON object. Repeat for multiple attachments.
  • --dry-run — Preview without posting.
  • --force — Confirm this external write.
  • --json

sb teams status

Show Microsoft Teams connector configuration status.

Flags:

  • --json

sb teams update-planner-task

Update a Microsoft Planner task using an ETag-safe write.

Arguments:

  • task_id (required)

Flags:

  • --etag — Planner task ETag. Fetched when omitted.
  • --title — New task title.
  • --bucket-id — New bucket ID.
  • --due-date-time — New due timestamp.
  • --start-date-time
  • --percent-complete
  • --priority
  • --assign
  • --dry-run — Preview without updating.
  • --force — Confirm this external write.
  • --json

sb teams update-planner-task-details

Update Planner task description, preview, checklist, or references.

Arguments:

  • task_id (required)

Flags:

  • --etag — Planner task details ETag.
  • --description
  • --preview-type
  • --checklist-item
  • --checklist-json
  • --references-json
  • --dry-run — Preview without updating.
  • --force — Confirm this external write.
  • --json

sb templates

Shortcut for sb tasks templates.

sb today

What matters today — daily operator briefing from Antahkarana and local work state.

sb tool-sets

List available modular tool sets and profile defaults.

sb tools

List chat tool availability under current permission + web modes.

sb tools-audit

Audit tool registry against the 7 canonical Data Agent primitives.

sb trace

Quick trace lookup: show a run record by ID (alias for sb traces show).

sb traces

Decision trace inspection and replay.

sb traces agent-list

List durable agent-run trace receipts (separate from decision traces).

Flags:

  • --limit
  • --json — Output JSON

sb traces agent-show

Show a durable agent-run trace receipt (separate from decision traces).

Arguments:

  • trace_id (required)

Flags:

  • --json — Output JSON

sb traces approvals

Show approval records attached to a trace and its linked governed writes.

Arguments:

  • trace_id

Flags:

  • --session-id — Resolve the latest trace for this session when trace_id is omitted.
  • --governed — When resolving by session, only consider traces with institutional governance artifacts.
  • --institutional-mode — When resolving by session, only consider traces in this institutional mode.
  • --decision — Filter approval outcome: approved, denied, requested.
  • --json — Output JSON

sb traces challenges

Show the persisted institutional challenge report for a trace.

Arguments:

  • trace_id

Flags:

  • --session-id — Resolve the latest trace for this session when trace_id is omitted.
  • --governed — When resolving by session, only consider traces with institutional governance artifacts.
  • --institutional-mode — When resolving by session, only consider traces in this institutional mode.
  • --json — Output JSON

sb traces correlation

Correlate one trace id across decision trace, events, and tool calls.

Arguments:

  • trace_id

Flags:

  • --session-id — Resolve the latest trace for this session when trace_id is omitted.
  • --governed — When resolving by session, only consider traces with institutional governance artifacts.
  • --institutional-mode — When resolving by session, only consider traces in this institutional mode.
  • --limit-events
  • --limit-tool-calls
  • --institutional-summary — Inline the consolidated institutional audit envelope.
  • --include-payloads — Include decoded event/tool payloads in table output.
  • --json — Output JSON

sb traces coverage

Show trace-id coverage across runtime events and tool calls.

Flags:

  • --limit-event-types — Max event types in coverage breakdown.
  • --json — Output JSON

sb traces events

List canonical event-log rows with runtime trace metadata.

Flags:

  • --limit — Max events to return
  • --event-type — Filter by event type
  • --session-id — Filter by session id
  • --status — Filter by status
  • --trace-id — Filter by trace id
  • --include-payloads — Include inputs/outputs/artifacts payloads in table output.
  • --json — Output JSON

sb traces gate

Evaluate release gate recommendation from recent decision traces.

Flags:

  • --limit — Traces to evaluate
  • --json — Output JSON

sb traces ids

List available trace ids merged from decision traces, events, and tool calls.

Flags:

  • --limit — Max trace ids to return
  • --session-id — Filter by session id
  • --decision-status — Filter decision trace status
  • --governed — Only include traces with institutional governance artifacts.
  • --institutional-mode — Filter governed traces by institutional mode.
  • --event-type — Filter event source by event type
  • --tool-name — Filter tool-call source by tool name
  • --json — Output JSON

sb traces institutional-summary

Show one consolidated institutional audit envelope for a trace.

Arguments:

  • trace_id

Flags:

  • --session-id — Resolve the latest trace for this session when trace_id is omitted.
  • --governed — When resolving by session, only consider traces with institutional governance artifacts.
  • --institutional-mode — When resolving by session, only consider traces in this institutional mode.
  • --json — Output JSON

sb traces list

List decision traces.

Flags:

  • --limit — Max traces to return
  • --task-graph-id — Filter by task graph
  • --session-id — Filter by session id
  • --status — Filter by status
  • --governed — Only include traces with institutional governance artifacts.
  • --institutional-mode — Filter governed traces by institutional mode.
  • --json — Output JSON

sb traces memory-proposals

Show persisted governed memory-review proposals linked to a trace.

Arguments:

  • trace_id

Flags:

  • --session-id — Resolve the latest trace for this session when trace_id is omitted.
  • --governed — When resolving by session, only consider traces with institutional governance artifacts.
  • --institutional-mode — When resolving by session, only consider traces in this institutional mode.
  • --status — Filter by review status.
  • --limit — Max proposals to return
  • --json — Output JSON

sb traces policy-decisions

Show persisted institutional policy decisions for a trace.

Arguments:

  • trace_id

Flags:

  • --session-id — Resolve the latest trace for this session when trace_id is omitted.
  • --governed — When resolving by session, only consider traces with institutional governance artifacts.
  • --institutional-mode — When resolving by session, only consider traces in this institutional mode.
  • --status — Filter by institutional decision status: allow, deny, require_approval.
  • --json — Output JSON

sb traces replay

Replay a stored trace into an orchestrator-like response envelope.

Arguments:

  • trace_id (required)

Flags:

  • --json — Output JSON

sb traces run-graph

Show the persisted institutional run graph for a trace.

Arguments:

  • trace_id

Flags:

  • --session-id — Resolve the latest trace for this session when trace_id is omitted.
  • --governed — When resolving by session, only consider traces with institutional governance artifacts.
  • --institutional-mode — When resolving by session, only consider traces in this institutional mode.
  • --json — Output JSON

sb traces show

Show a single decision trace envelope.

Arguments:

  • trace_id

Flags:

  • --session-id — Resolve the latest trace for this session when trace_id is omitted.
  • --governed — When resolving by session, only consider traces with institutional governance artifacts.
  • --institutional-mode — When resolving by session, only consider traces in this institutional mode.
  • --institutional-summary — Inline the consolidated institutional audit envelope.
  • --json — Output JSON

sb traces tool-calls

List chat tool-call rows with runtime trace metadata.

Flags:

  • --limit — Max rows to return
  • --session-id — Filter by session id
  • --trace-id — Filter by trace id
  • --tool-name — Filter by tool name
  • --status — Filter by status
  • --json — Output JSON

sb travel

Cognitive travel planning operations.

sb travel ask

Answer travel questions with ranked recommendations and rationale.

Arguments:

  • query (required)

Flags:

  • --origin — Optional origin city/airport
  • --destination — Optional destination city/airport
  • --dates, -d — Start,end dates (YYYY-MM-DD,YYYY-MM-DD)
  • --budget, -b — Budget amount
  • --max-results — Maximum recommendation rows
  • --json — Emit JSON output

sb travel book

Mock-book a travel plan (requires --confirm).

Arguments:

  • goal_id (required)
  • plan_id (required)

Flags:

  • --confirm — Confirm booking (required)
  • --json — Emit JSON output

sb travel chat

Interactive unified travel planner chat mode (OTA + cognitive).

Flags:

  • --user-id — Traveler profile id
  • --agentic-mode — Combined planner mode: auto|off|force.
  • --debug-sections — Include Content/Context/Composition/Consideration sections.
  • --confirm-side-effects — Allow booking/change/cancel side effects for this chat session.
  • --consent-save-prefs — Allow saving stable travel preferences into local travel vault.
  • --json — Emit JSON output

sb travel cognitive

Multi-agent cognitive travel planning with Think → Plan → Act → Reflect.

Arguments:

  • query (required)

Flags:

  • --origin — Optional origin city/airport
  • --destination — Optional destination city/airport
  • --dates, -d — Start,end dates (YYYY-MM-DD,YYYY-MM-DD)
  • --budget, -b — Budget amount
  • --max-results — Maximum recommendation rows
  • --view — View preset: user|debug|ops|audit
  • --show — Extra sections (comma-separated)
  • --hide — Sections to hide (comma-separated)
  • --trace — Show composition trace
  • --json — Emit JSON output

sb travel decisions

List persisted travel semantic decision records.

Flags:

  • --type — Optional decision type filter
  • --limit — Maximum records
  • --json — Emit JSON output

sb travel evolve

Mind Evolution + RCC search for travel tasks.

sb travel evolve demo

Run deterministic toy travel evolution demo end-to-end.

Flags:

  • --output-dir — Optional output directory for run artifacts.
  • --dump-pop — Write full population dump.
  • --json — Emit JSON output

sb travel evolve run

Run travel Mind Evolution on a JSON task using selected evaluator.

Flags:

  • --task — Path to task JSON file.
  • --evaluator — Evaluator class in module:Class format.
  • --llm-backend — LLM backend: default|toy.
  • --output-dir — Optional output directory for run artifacts.
  • --dump-pop — Write full population dump.
  • --n-gens
  • --n-island
  • --n-convs
  • --n-seq
  • --n-reset-interval
  • --n-reset
  • --n-top
  • --n-candidate
  • --n-parent
  • --pr-no-parents
  • --n-emigrate
  • --n-retries
  • --selection-temperature
  • --reset-strategy
  • --population-cap
  • --serial — Run island conversations serially.
  • --random-seed
  • --json — Emit JSON output

sb travel goal

Show goal detail and event history.

Arguments:

  • goal_id (required)

Flags:

  • --json — Emit JSON output

sb travel goals

List travel goals.

Flags:

  • --status, -s — Filter by status
  • --json — Emit JSON output

sb travel intents

List all supported travel intents and metadata.

Flags:

  • --json — Emit JSON output

sb travel kg-query

Query the travel knowledge graph.

Flags:

  • --query, -q — Query string (e.g. 'route:DEL:GOI')
  • --json — Emit JSON output

sb travel kg-seed

Seed the travel knowledge graph with airports, routes, and hotels.

Flags:

  • --json — Emit JSON output

sb travel maps-ground

Ground a location query against Google Maps via Gemini API.

Arguments:

  • query (required)

Flags:

  • --lat — Latitude for location context
  • --lng — Longitude for location context
  • --widget — Request a Maps widget context token
  • --context — Non-PII preference context
  • --force — Force grounding even if geo intent not detected
  • --json — Emit JSON output

sb travel meditate

Antahkarana-powered travel — runtime responsibility loop + LLM enrichment.

Arguments:

  • query (required)

Flags:

  • --origin — Optional origin city/airport
  • --destination — Optional destination city/airport
  • --dates, -d — Start,end dates (YYYY-MM-DD,YYYY-MM-DD)
  • --budget, -b — Budget amount
  • --max-results — Maximum recommendation rows
  • --provider, -p — LLM provider for layer enrichment
  • --model, -m — LLM model for layer enrichment
  • --show-layers — Show per-layer insights
  • --trace — Show full impression trace
  • --json — Emit JSON output

sb travel midway

Recommend midpoint stopover stays for known road corridors.

Arguments:

  • origin_city (required)
  • destination_city (required)

Flags:

  • --dates, -d — Check-in/out dates (YYYY-MM-DD,YYYY-MM-DD)
  • --budget, -b — Per-stopover budget cap
  • --min-stars — Minimum hotel stars
  • --max-results — Maximum recommendations
  • --json — Emit JSON output

sb travel monitor

Check active trips for disruptions.

Flags:

  • --json — Emit JSON output

sb travel plan

Plan travel from natural language (unified mode) or run legacy destination planning.

Arguments:

  • user_text (required)

Flags:

  • --origin, -o — Origin airport code
  • --destination — Optional destination override
  • --dates, -d — Start,end dates (YYYY-MM-DD,YYYY-MM-DD)
  • --budget, -b — Budget amount
  • --currency — Budget currency
  • --agentic-mode — Combined planner mode: auto|off|force.
  • --debug-sections — Include Content/Context/Composition/Consideration sections.
  • --confirm-side-effects — Allow booking/change/cancel side effects for this request.
  • --consent-save-prefs — Allow saving stable travel preferences into local travel vault.
  • --json — Emit JSON output

sb travel preferences

List travel preferences.

Flags:

  • --category, -c — Filter by category
  • --json — Emit JSON output

sb travel recommend

Show top recommended plan after simulation.

Arguments:

  • goal_id (required)

Flags:

  • --json — Emit JSON output

sb travel semantic

Show semantic runtime diagnostics (content/context/composition/consideration).

Arguments:

  • query (required)

Flags:

  • --origin — Optional origin city/airport
  • --destination — Optional destination city/airport
  • --dates, -d — Start,end dates (YYYY-MM-DD,YYYY-MM-DD)
  • --budget, -b — Budget amount
  • --max-results — Maximum recommendation rows
  • --json — Emit JSON output

sb travel simulate

Run Monte Carlo simulation on candidate plans.

Arguments:

  • goal_id (required)

Flags:

  • --runs, -n — Number of Monte Carlo runs
  • --json — Emit JSON output

sb ui-schema

Generate and print/export the CLI schema for UI command parity.

sb vault

Vault management — stats, cleanup, and retitling.

sb vault clean

Clean vault: archive duplicate/re-ingested files and retitle hash-named files.

Flags:

  • --apply — Actually rename/archive (default is dry-run)
  • --verbose, -v — Show all files including clean ones

sb vault stats

Show vault file counts and detected problems (read-only).

sb vault sync-people

Seed the people graph from vault frontmatter and USER.md.

Flags:

  • --verbose, -v — Show each person upserted

sb verify

Run deterministic verification checks for a plan, run, or task.

sb voice-test

Record from your microphone and transcribe with whisper-cli.

sb watch

Watch a drop folder and ingest supported files.

sb week

Weekly synthesis: emerging thesis, contradictions, knowledge gaps.

sb wiki

LLM Wiki — persistent, compounding personal knowledge base.

sb wiki compare

Compare two wiki topics using graph context and LLM synthesis.

Arguments:

  • slug_a (required)
  • slug_b (required)

Flags:

  • --provider, -p — LLM provider name
  • --model, -m — LLM model name
  • --json — Output JSON

sb wiki graph

Analyse the wiki as a knowledge graph.

Flags:

  • --god-nodes, -g — Number of god nodes to show
  • --json — Output JSON

sb wiki ingest

Ingest a source document or URL into the wiki.

Arguments:

  • source (required)

Flags:

  • --title, -t — Override source title
  • --provider, -p — LLM provider name
  • --model, -m — LLM model name
  • --json — Output JSON

sb wiki lint

Health-check the wiki: orphans, missing pages, contradictions, gaps.

Flags:

  • --fix — Apply safe repairs after linting (LLM-planned when a provider is available).
  • --no-llm — Skip LLM-based checks (structural only)
  • --provider, -p — LLM provider name
  • --model, -m — LLM model name
  • --json — Output JSON

sb wiki list

List all pages in the wiki.

Flags:

  • --type, -t — Filter by type: entity|concept|source|synthesis|answer
  • --json — Output JSON

sb wiki new

Create a blank wiki page manually.

Arguments:

  • title (required)

Flags:

  • --type, -t — Page type: entity|concept|synthesis
  • --tags — Comma-separated tags

sb wiki query

Answer a question by searching and synthesising wiki pages.

Arguments:

  • question (required)

Flags:

  • --file, -f — File answer back into the wiki
  • --provider, -p — LLM provider name
  • --model, -m — LLM model name
  • --json — Output JSON

sb wiki show

Show a wiki page.

Arguments:

  • slug (required)

Flags:

  • --raw — Show raw markdown (no rendering)

sb wiki stats

Show wiki statistics.

Flags:

  • --json — Output JSON

sb wiki synthesize

Generate a synthesis page bridging the two most distinct wiki communities.

Flags:

  • --provider, -p — LLM provider name
  • --model, -m — LLM model name
  • --json — Output JSON

sb workflow

Workflow automation operations.

sb workflow automations

Recurring workflow-native automations.

sb workflow automations clear-invoke-token

Remove per-automation invoke token (only global serve auth can run the automation).

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb workflow automations create

Create a recurring automation that materializes scheduled tasks.

Arguments:

  • name (required)

Flags:

  • --prompt — Prompt or task text to materialize
  • --schedule — Schedule: interval: or cron:<5-field>
  • --priority — Task priority: low|medium|high|critical
  • --project — Associated project ID
  • --paused — Create in paused state
  • --generate-invoke-token — Create a per-automation token for POST .../run (shown once; also rotatable via API/CLI).
  • --json — Output JSON

sb workflow automations delete

Delete one recurring automation definition.

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb workflow automations list

List recurring workflow-native automations.

Flags:

  • --status — Filter by active|paused
  • --json — Output JSON

sb workflow automations pause

Pause one automation.

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb workflow automations resume

Resume one automation.

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb workflow automations rotate-invoke-token

Rotate the per-automation invoke token used for narrow API access to run.

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb workflow automations run

Run one automation now or materialize all currently due automations.

Arguments:

  • automation_id

Flags:

  • --due-only — Run all due automations.
  • --as-of — Override current timestamp.
  • --dry-run — Preview without creating tasks.
  • --limit — Maximum due automations to run.
  • --json — Output JSON

sb workflow automations show

Show one recurring automation definition.

Arguments:

  • automation_id (required)

Flags:

  • --json — Output JSON

sb workflow disable

Disable a workflow in SQLite state.

Arguments:

  • workflow_id (required)

sb workflow enable

Enable a workflow in SQLite state.

Arguments:

  • workflow_id (required)

sb workflow lint

Lint workflow specs.

Flags:

  • --path — Workflow file or directory
  • --json — Output JSON

sb workflow list

List workflow specs from committed + overlay paths.

Flags:

  • --all — Include disabled workflows
  • --json — Output JSON

sb workflow run

Run one workflow spec by id.

Arguments:

  • workflow_id (required)

Flags:

  • --input — Input key=value
  • --input-json — JSON file with workflow inputs
  • --session-id — Optional session id for trace lineage
  • --json — Output JSON

sb workflow schedule

Schedule a task with optional recurrence.

Arguments:

  • title (required)

Flags:

  • --due — Due date (ISO or relative: +1d, +1w)
  • --recurrence — Recurrence: daily|weekly|monthly
  • --priority, -p — Priority: low|medium|high|critical
  • --project — Associated project ID

sb workflow show

Show one workflow spec.

Arguments:

  • workflow_id (required)

Flags:

  • --json — Output JSON

sb workflow tasks

List scheduled tasks.

Flags:

  • --status, -s — Filter by status
  • --project — Filter by project

sb workforce

AI workforce orchestration operations.

sb workforce capabilities

Show workforce authoring capabilities.

Flags:

  • --json — Output JSON

sb workforce continue

Continue a dependency-aware detached workforce run.

Arguments:

  • run_id (required)

Flags:

  • --detach, --no-detach — Launch newly ready roles in background
  • --json — Output JSON

sb workforce create

Create a local overlay workforce from a reviewed spec file.

Flags:

  • --file — YAML/JSON workforce spec
  • --json — Output JSON

sb workforce dashboard

Show workforce run and lesson dashboard state.

Flags:

  • --workforce — Filter by workforce id
  • --json — Output JSON

sb workforce deploy

Create or update the recurring session automation for a workforce.

Arguments:

  • workforce_id (required)

Flags:

  • --json — Output JSON

sb workforce disable

Disable a workforce in the local registry.

Arguments:

  • workforce_id (required)

Flags:

  • --json — Output JSON

sb workforce enable

Enable a workforce in the local registry.

Arguments:

  • workforce_id (required)

Flags:

  • --json — Output JSON

sb workforce lessons

Workforce lesson review operations.

sb workforce lessons list

List workforce lessons.

Flags:

  • --workforce — Filter by workforce id
  • --status — Filter by status
  • --json — Output JSON

sb workforce lessons promote

Promote a workforce lesson into the generated lessons register.

Arguments:

  • lesson_id (required)

Flags:

  • --json — Output JSON

sb workforce lessons review

Mark a workforce lesson reviewed.

Arguments:

  • lesson_id (required)

Flags:

  • --json — Output JSON

sb workforce lint

Lint workforce specs.

Flags:

  • --path — Workforce file or directory
  • --json — Output JSON

sb workforce list

List workforce specs from committed + overlay paths.

Flags:

  • --all — Include disabled workforces
  • --json — Output JSON

sb workforce proposals

Workforce review proposal operations.

sb workforce proposals list

List workforce review proposals.

Flags:

  • --workforce — Filter by workforce id
  • --status — Filter by status
  • --kind — Filter by proposal kind
  • --json — Output JSON

sb workforce proposals reject

Reject a workforce proposal.

Arguments:

  • proposal_id (required)

Flags:

  • --json — Output JSON

sb workforce proposals review

Mark a workforce proposal reviewed.

Arguments:

  • proposal_id (required)

Flags:

  • --json — Output JSON

sb workforce run

Run one workforce by id.

Arguments:

  • workforce_id (required)

Flags:

  • --project — Limit run to a project id
  • --dry-run — Preview without launching roles
  • --detach, --no-detach — Run roles in background
  • --json — Output JSON

sb workforce show

Show one workforce spec.

Arguments:

  • workforce_id (required)

Flags:

  • --json — Output JSON

sb workforce templates

List workforce authoring templates.

Flags:

  • --json — Output JSON

sb workforce validate

Validate a user-authored review-first workforce spec.

Flags:

  • --file — YAML/JSON workforce spec
  • --json — Output JSON

sb workspace

Multi-tenant workspace registry — workspaces and bearer tokens (Phase 3).

sb workspace create

Register a new workspace.

Arguments:

  • slug (required)

Flags:

  • --display-name — Human-readable display name (defaults to the slug).
  • --json — Emit machine-readable JSON instead of a table.

sb workspace list

List all registered workspaces.

Flags:

  • --json — Emit JSON.

sb workspace token

Workspace bearer token operations.

sb workspace token issue

Issue a fresh bearer token. The raw token is shown ONCE.

Flags:

  • --workspace, -w — Workspace slug to bind the token to.
  • --actor — Optional allowed-actor-id. Pass once per actor. When set, the x-sb-actor-id header is required and must match one of these.
  • --scope — Optional scope tag (free-form, repeatable).
  • --json — Emit JSON.

sb workspace token list

List tokens (hashes only — raw values are never recoverable).

Flags:

  • --workspace, -w — Filter by workspace slug.
  • --json — Emit JSON.

sb workspace token revoke

Revoke a token by id.

Arguments:

  • token_id (required)

sb workspace-context

Workspace context — standing instructions auto-prepended to every bridge call.

sb workspace-context clear

Delete the workspace-context file.

Flags:

  • --yes, -y — Skip confirmation.

sb workspace-context edit

Open the workspace-context file in $EDITOR (creates it if missing).

Flags:

  • --editor, -e — Editor binary (defaults to $VISUAL, $EDITOR, or vi).

sb workspace-context path

Print the resolved workspace-context file path.

sb workspace-context set

Overwrite the workspace-context file with the provided content.

Arguments:

  • content (required)

sb workspace-context show

Show the current workspace-context content (or the rendered block).

Flags:

  • --rendered — Show the rendered block as it is injected into bridge prompts.

sb workspace-context status

Summarise whether workspace context is active, where it lives, how big it is.