Command
Run the full bureau maintenance lifecycle in one pass — compile pending minutes into dossiers, scan the recursion engine for changed claims, lint for consistency, surface everything that needs human review (by trust tier AND freshness), and rebuild the gazette. Use when running bureau:cycle, or when the user asks to…
Command
Write the rich minute for the current session (intent, decisions, changes, open threads).
Command
Show the pre-change blast radius of a dossier — which pages (transitively) rest on its claim, so you can see the review cost before you touch it. Use when running bureau:impact, or when the user asks "what depends on this?", "what breaks if I change X?", "who rests on this claim?", or "is this safe to edit?".
Command
Scaffold a bureau workspace (cabinets + logbook) in the current repo and wire the press.
Command
Build the gazette (the offline gazette) with the bundled press, and open it for inspection.
Command
Sweep the cabinets for semantic inconsistencies (contradictions, superseded claims, gaps, drift) and write a findings report.
Command
Take a live note — summarize the current session's decisions/open threads into the running minute (or ephemerally if no workspace).
Command
Pair this repo's external bureau workspace (its .bureau-id) to a local path on this machine.
Command
Answer a question from the bureau canon — cited, tier-aware, and never stating an unverified claim as fact.
Command
The human double-check gate — review AI-written cabinet claims and promote the vetted ones to canonical, reject the rest.
Command
Open the chamber — a localhost-only room serving the gazette read-only plus an intake form that files proposed claims as append-only logbook minutes (never canon). The interactive write surface that feeds the gate.
Command
Version the canon with git — pin a named, reproducible snapshot, diff two versions semantically, or render the board as it was at any past commit. Use when running bureau:snapshot, or when the user asks to snapshot / tag / freeze / pin a version of the canon, compare two versions, see what changed, or view the canon…
Command
Report the state of the bureau — uncompiled sessions, pages by trust tier, dependency freshness (needs-review/stale from the recursion engine), and what needs your attention.
Hook
Runs when a session ends, executing capture-stub.mjs via node. From xiaolai/bureau.
Hook
Runs when a session starts on compact, executing scribe-checkpoint.mjs via node. From xiaolai/bureau.
Skill Claude CodeCodex
Record an architecture decision as a MADR ADR page in the canon. Use when running bureau:adr, or when the user asks to write/record an ADR or an architecture decision, or to supersede a prior decision. Authors a proposed page only — never approves.
Skill Claude CodeCodex
Prepare a human bulk-approval of the whole approvable review backlog — show what would be promoted to canonical, take an in-session confirmation, then hand the human the single pre-filled gazette approve --all --by human line to fire themselves. Use when running bureau:approve-all, or when the user asks to approve…
Skill Claude CodeCodex
Write a faithful, append-only minute for an AI session — intent, decisions, changed files, open threads, and a transcript pointer. Use when filing a session into a bureau workspace (e.g. via bureau:file-session), or when the user asks to record/minute the current session.
Skill Claude CodeCodex
Let Codex (via cc-suite) deliberate over the bureau review queue as the user's representative — read each unapproved claim, recommend approve or hold, and either hand the human the exact approve command (advisor) or, only where the workspace opted the codex authority in, commit under --by codex (delegate). Use when…
Skill Claude CodeCodex
Distil minutes into consistency-checked dossiers — the canonical SSOT — writing provenance links back to the sessions that introduced each claim. Use when running bureau:compile, or when the user asks to turn the logbook into canon / update the cabinets / build the knowledge base from sessions.
Skill Claude CodeCodex
Orientation for the bureau plugin — the records-office model, the capture → compile → review gate, the trust tiers, the crew, and which bureau: command serves which intent. Read this FIRST before using any bureau: command, or whenever working in a repo that has a bureau workspace (a canon/ directory + a BUREAU.md), so…
Skill Claude CodeCodex
Sweep the dossiers for semantic inconsistencies that the press's structural check cannot see — free-text contradictions, superseded claims, undocumented gaps, and vocabulary drift. Use when running bureau:lint, before a milestone, or when the user asks to check the canon for contradictions / consistency / drift.
Skill Claude CodeCodex
Answer from the bureau canon with citations and trust tiers — read dossiers as memory, honor each page's status, and never present an unverified claim as fact. Use when running bureau:query, or when the user asks what the project knows / what was decided / what the canon says about something.
Skill Claude CodeCodex
The human double-check gate for AI-written memory. Show every cabinet claim not yet approved — with its provenance and automatic check result — and let the human promote it to canonical or reject it. Use when running bureau:review, or when the user asks to approve / vet / sign off on what the AI wrote to memory before…