arbeitsplan-statuslisted

Read-only report on an arbeitsplan run that has already started — phases done, candidates built and refuted, referee verdicts, open doubts, halts with their reasons, dispatch budget used, whether a lock is open, and the single next command. Use when the user asks how the run is going, what the swarm decided, why it stopped, how to resume it, why something was denied, or what is still holding a lock. Advances nothing and writes nothing.
Anselmoo/werkstoff · ★ 1 · AI & Automation · score 67

Install: claude install-skill Anselmoo/werkstoff

# Status Reports what has already run. It has nothing to say outside its own pipeline, and it never advances one. ## Steps 1. **Find the runs.** `analysis/arbeitsplan/*/workflow.json`. Report the most recent unless the user names one. 2. **Read the record, not the chat.** The run directory is the plan of record: `workflow.json` is the plan and never changes; `run.jsonl` is what happened. Ask the script, which derives everything from the log every time and stores no counter: ```bash python3 "${CLAUDE_PLUGIN_ROOT}/scripts/record_event.py" status --run <runId> ``` It reports closed and open phases, a halt with its reason, a pending plan-mode phase, a `FAILED-<stamp>.json` refusal, `complete.json`, counts per status (`refuted` candidates stay in the record after their worktrees are deleted; `doubt` entries carry `resolves_if`), and the **single next command**. Quote that command; do not compose your own. A run with no `run.jsonl` predates the record — say so, and fall back to the phase markers under `.takt/<runId>/`. 3. **Count the dispatch ledger** at `analysis/arbeitsplan/<runId>/dispatch/`. One file per distinct dispatch; the count against `budget.totalDispatches` is how much room is left. 4. **Report the lock.** If `run_scope.json` is open, say which phase it arms and how old it is. **Do not close it** — a stale lock denies every edit, which is the safe direction. 5. **Report candidate state** from the worktrees that still exist,