claude-org

Triage open Issues and have the Lead present "next-work candidates (N items + 1 recommendation)" to the human. Run the deterministic tool tools/work_discovery_scan.py once and stop after rendering the candidate JSON in the human-readable format defined by design doc §5.2 (propose-only). The trigger is restricted to the Lead. Manual / event-driven only (no resident /loop). Fires when the Lead manually invokes for "show next-work candidates", "triage", "what's next?", or for proactive next-dispatch after a PR merge.

Dispatch a Worker Claude to delegate work. The Lead is the command tower; hands-on execution is, in principle, left to Workers. Fires when a user request involves actual hands-on work such as file edits, implementation, or investigation.

Start up the org. Load the previous state and brief, then launch the dispatcher pane. Run this once right after starting Claude Code. Also triggered by "start", "boot", "begin", etc. The curator is not launched (it has moved to on-demand launch by the dispatcher when the threshold check at worker close fires).

Suspend the org and fully stop it, down to and including the broker daemon. Use when told "stop it completely", "bring it down including the daemon", "org down it", "shut everything down". An integrated shutdown that runs suspend (state save + stopping ja-managed processes / panes) to completion and then calls `claude-org-runtime org down`. For a plain "suspend" / "done for today", use /org-suspend, which leaves the daemon running.

After the user approves a worker's completion report, the Lead handles push / PR creation / CI monitoring / review-feedback loop / final close after PR merge. Triggers: (1) immediately after a worker submits a completion report and the user gives explicit approval such as "OK" or "go ahead", (2) when review feedback / CI failure arrives on the GitHub PR and you need to send fix instructions back to the worker, (3) when the PR is merged and the final close conditions are met. The initial "delegate work to a worker" step is handled by org-delegate, not by this skill.

Suspend the org and persist all state to disk. Use when the user says "suspend", "save and exit", "I want to close", "let's stop for now", "done for today".

Run PR CI / merge monitoring (tools/pr-watch.sh) in a dedicated pane pr-watch-<PR> inside the broker tmux session. When the Lead starts it with `/pr-watch-pane <PR>` right after creating a PR, the monitoring runs in ja-root cwd, outside the sandbox, and continues independently of /clear or the Lead session's lifetime. Idempotent start via the pane name (no double monitoring), identity registration with role=watcher, and the pane auto-closes when monitoring ends (CI green / PR merged / timeout). The Bash tool's background mode is session-lifetime-bound and unsuitable for a long-running watcher, so this skill is the recommended path. Triggered by "watch CI in a pane", "run pr-watch in a pane", "watch PR <N>'s CI", etc.

Read the handover file written by /dispatcher-handover and bring the Dispatcher back in a fresh session. Use it on the very first turn after /clear. Atomically update `dispatcher_pane_id` / `dispatcher_peer_id` in state.db, and resume the `/loop 3m` worker monitoring loop. Use when the Secretary instructs "resume the Dispatcher" / "resume" / "pick up from the handover". This is not /org-start (the Worker / Secretary / Curator are assumed to still be alive).

Stop the attention watcher pane started by `/org-attention-start`. It identity-checks the pane_id recorded in `.state/attention_pane.json` against the name/role from `list_panes`, and destroys the pane via `close_pane` (pane destroy) only when it still points at the attention watcher (if the pane_id has already been reassigned to another pane, it does not close it and deletes it as a stale sidecar). Triggered by "stop attention", "halt notification monitoring", "shut down the watcher", etc.

A belt conveyor in which the Lead receives an "outline of the approved scope" from the human and, inside it, self-drives triage → worker dispatch → iteration → verify → push → PR creation → CI watch in a completion-driven way, always stopping at the merge gate for every PR. A thin orchestrator that calls /work-discovery + /org-delegate + verify + /org-pull-request + /org-escalation. Fires when the human explicitly gives a scope approval saying "this range may self-drive" and wants to flow multiple candidates through sequentially in a completion-driven way (e.g. "self-drive the top S-class bug fixes from triage for as many free panes as there are; I'll decide each merge"). The trigger is restricted to the Lead. Whenever it touches an out-of-scope candidate, a judgment boundary, or an exit condition, it must halt. Merge is never automated.

Skill inventory pass (deprecation / consolidation / owner-missing checks). Fires on a state-based trigger only: when the candidate queue knowledge/skill-candidates.md has 5 or more pending entries, OR when the work-skill count under .claude/skills/ (excluding org-*) reaches 20 or more. Not started by a time-based /loop (avoids polluting raw logs on quiet days).

To avoid continuing the Dispatcher session while its context stays bloated, write the monitoring state (active workers / latest polling cursor / pending escalations) to a handover file, then — on the Secretary's instruction — prepare a fresh Dispatcher session via /clear → /dispatcher-resume. Use when a DISPATCHER_HANDOVER peer message arrives from the Secretary, or when the Dispatcher itself judges that context has grown long.

Consolidate and reorganize accumulated raw learnings (knowledge/raw/). Called exactly once by a curator that the dispatcher spawned on demand when the threshold check at worker close (tools/check_curate_threshold.py) fired (the resident /loop is retired). Also fires manually when asked to "organize the knowledge".

When a worker sends a peer message such as "judgment escalation", "approval request", "OK to continue?", "scope-expansion proposal", "unexpected", "runbook deviation", or "block / blocker", the Lead does NOT pre-approve and instead escalates to the human. Triggers: receipt of a judgment-escalation / scope-expansion / blocker message from a worker. Normal progress / completion reports are handled by org-delegate Step 5 (1) / (2a). This skill owns the canonical "do not approve on your own; update the registers" flow.

Retrospective on the delegation process. When a worker delegation finishes, look back at how the delegation itself went and record process-improvement learnings. Also decide whether the completed task's pattern should be promoted to a work-skill. Technical retrospectives on the actual work are done by the worker automatically and are not handled here.

Load the handover file written by /secretary-handover and resume the Secretary in a fresh session. Use it on the very first turn after /clear. Trigger phrases: "resume the Secretary", "resume", "pick up from the handover". This is not /org-start (Dispatcher / Curator are assumed to still be alive).

Start the attention notification watcher (active OS notification + sound for awaiting approval / awaiting decision / CI failures, etc.) resident as a vertical split on the right side of the dispatcher pane. If `.state/attention.json` is not in place yet, the ja default template is auto-copied from `tools/templates/attention.example.json`. After startup, the pane id is recorded in `.state/attention_pane.json` and consumed by `/org-attention-stop`. Triggered by "start attention", "begin notification monitoring", "spin up the watcher", etc. `/org-start` does **not** auto-start it (explicit-start policy).

Read-only skill that **only emits commands** for the human to use to **directly attach via tmux** to live org panes (secretary / dispatcher / worker). It joins `mcp__org-broker__list_panes` (logical pane list) with `/usr/bin/tmux -L claude-org-broker list-panes -a` (pane_id <-> session mapping) on pane_id (%N), and prints an attach command for each pane labeled with role + name(task_id) (read-only `-r` / writable without `-r` / detach with `Ctrl-b d`). Triggered by "I want to see a worker's pane", "I want to attach to a worker pane", "I want to peek at the dispatcher directly", "let me into a pane", "tell me the attach command", "connect via tmux", "enter the worker's screen", etc. Does not attach itself or change any pane (it only displays command strings and a pane list). Does NOT fire for delegation (org-delegate), dashboard listing (org-dashboard), or watcher stop (org-attention-stop).

Place and update the Claude Code permission settings and environment variables required by every role in the org (Lead / Dispatcher / Curator / Worker), in one go. Triggers: "set up", "update permissions", "do the setup", "permissions config", "org-setup", etc.

Shared skill that decides whether a work pattern should be promoted to a skill. Called from org-retro and org-curate; returns a 3-value verdict (skill_recommend / candidate_queue / curated_only) with rationale. Does not auto-create skills. A "recommend" appends to knowledge/skill-candidates.md, and the Lead later batch-asks the human once the queue accumulates — a two-stage handoff.

To avoid continuing a session while the Secretary's context stays bloated, write recent interactions, in-flight work, and org state to a handover file, then prepare to start a fresh Secretary session via /clear → /secretary-resume. Use when the user says "refresh", "I want to hand over the Secretary", or "clean up the context", or when the Secretary itself judges that context has grown long.

Resume a suspended org. Use when `.state/org-state.md` exists with Status: SUSPENDED, or when the user says "resume", "pick up from where we left off", or "where did we stop last time?". Also handles the automatic startup briefing.

Refresh the org dashboard and open it in a browser. Triggers: "show me the dashboard", "visualize the situation", "show me the project list", "show me the big picture", etc.