uncovertechtalent/cto-powersuit / plugins/ai-ops/skills/coordinator
coordinator
VerifiedRun a coordinator session that routes work to worker sessions or subagents and verifies their artifacts, with all state in files: a ledger of tasks, briefs that name files and done criteria, caps on parallel workers, and resume after a stop. Use when one person runs several agent sessions at once, when a batch of tasks is too big for one session, or after a session limit cut work off. Activates on "/coordinator", "coordinate these sessions", "split this across agents", "resume the batch", "run a batch of tasks".
Coordinator
One session plans, briefs, routes and verifies. Workers do the work. Everything the coordinator knows lives in files, so a stopped session resumes from the ledger and a person can read the state at any time.
Rules of the pattern
- State in files.
work/ledger.mdholds every task: id, task, worker, state, artifact, verification, time. The coordinator writes it before it starts a worker and after it checks the result. - The coordinator routes and verifies. It writes briefs and checks artifacts. It does the work itself only when the task is smaller than the brief would be.
- Briefs are complete. Each brief names the goal, the files the worker may touch, the files it must read, what done looks like, the checks to run, and what to report. A worker has none of the coordinator's context.
- Verify the artifact. A worker's report says what it did; the file, the commit or the test output shows it. A task moves to
doneonly after the coordinator checks the artifact. - Caps. At most two or three workers run at once. Workers never start workers of their own; every brief says so. Two workers never write the same files.
- Resume. Every long task leaves a marker in the ledger (state
running, the worker, the start time). After a stop, read the ledger, check eachrunningtask's artifact, and set it todone,failedorqueuedbefore starting anything new.
Steps
- Set up. Copy
${CLAUDE_PLUGIN_ROOT}/templates/work/ledger.mdtowork/ledger.mdwithout overwriting. - Plan. Split the work into tasks a worker can finish alone. One row per task, state
queued. - Brief and start. For each task up to the cap: write the brief, set the row to
running, start the worker. - Verify. When a worker reports, open the artifact: read the diff, run the tests or the checks the brief named. Set
donewith the check in the "Verified by" column, orfailedwith the reason. - Integrate. The coordinator commits, merges or publishes, one task at a time, after verification. Workers do not commit unless the brief says so.
- Report. Summarise from the ledger: done, failed, blocked, and the evidence for each.
Output
The ledger, the briefs, the verified artifacts and the report.
Rules
- Never mark a task done from a report alone.
- Never let two workers write the same file.
- Never raise the cap or allow nested workers without the user's go.
Overlay
This skill works on the generic defaults in this file. Before you start, check for a private overlay:
[ -n "$POWERSUIT_OVERLAY" ] && ls -1 "$POWERSUIT_OVERLAY/ai-ops" 2>/dev/null || true
If $POWERSUIT_OVERLAY/ai-ops/ exists, read its README.md first, then the files in ai-ops/coordinator/ if that folder exists. Read other overlay files only when that README or this skill names them. Where the overlay and this file disagree, follow the overlay. If the variable is unset or the folder is missing, use the defaults and do not ask for an overlay. If the permission settings block the command, use the defaults and say in your output that the overlay was not checked. Never copy overlay content into this plugin, into its templates or into any public repository.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 469k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |