Skip to content

Heyvhuang/copybook-skills / standalone/coordinator/skills/coordinator

coordinator

Flagged

Coordinate an ongoing software delivery run when the user assigns this session the coordinator role or asks it to manage worker handoffs, gates, release approvals and progress. Use when starting, supervising or resuming that run, including a run with one active lane. Lanes run as Herdr workers (parallel-goal-fleet) on Claude or Codex, from whichever harness coordinates; T3 Code child tasks are the fallback. Do not start a coordinator run for discussion, a read-only review of the workflow, a standalone quick second opinion or an ordinary single-session edit.

FileKindSize
SKILL.mdskill md11.8 KB
references/example-run.mdmarkdown3.4 KB
references/herdr.mdmarkdown7.7 KB
references/lessons.mdmarkdown6.9 KB
references/models.mdmarkdown4.9 KB
references/pipeline.mdmarkdown6.1 KB
references/situations.mdmarkdown3.4 KB
references/t3.mdmarkdown4.9 KB
scripts/gemini.mjsexecutablescript2.3 KB
scripts/heavy.shexecutablescript829 B
scripts/status.shexecutablescript2.0 KB
scripts/watch-ready.shexecutablescript2.0 KB
templates/lane-brief.mdmarkdown1.5 KB
templates/release-packet.mdmarkdown1.4 KB
templates/run-record.mdmarkdown1.2 KB
templates/ship.mdmarkdown1.1 KB

references/example-run.md

View on GitHub
# Example run (3 days, about 160 PRs)

The production run this skill was distilled from: a public web app rebuilt from an older template (Next.js, hosted Postgres, a payments provider, an error tracker, preview and production builds on a hosting platform). One coordinator session in the Claude desktop app; Herdr workers for everything else. Run size: 107 lane briefs, 118 hand-overs, 166 gate train reports, 34 review files, a 1,484-line run record.

## Layout

`<repo>` is the shared checkout, `<work-dir>` the folder beside it holding worktrees and the run, `<run-dir>` = `<work-dir>/fleet`.

| Path | What |
| --- | --- |
| `<repo>` | shared main checkout; lanes never switch its branch or edit there |
| `<work-dir>/wt/<lane>` | one worktree per writer lane |
| `<run-dir>/STATE.md` | append-only log (TICK, READY, SHIPPED, APPLIED, USER YES, SHIP-STOP lines) |
| `<run-dir>/common.md` | rules every lane reads first: worktree, checks, git, build budget, speed rules, load rule, lane ownership |
| `<run-dir>/<lane>.md` | lane briefs |
| `<run-dir>/handover-<lane>.md` | hand-overs ending with the READY line |
| `<run-dir>/review-*.md` | Astra and Sol reviews |
| `<run-dir>/gate/queue.md`, `gate/train-N.md` | standing gate queue and reports |
| `<run-dir>/ship.md` | ship lane brief with the never-batch list |
| `<run-dir>/<release>-packet.md` | release packets (short and long form) |
| `<run-dir>/overnight-<date>.md` | overnight checklist with decisions; morning list |
| `<run-dir>/taste.md` | the user's taste notes, given to second readers |

## Standing lanes (a throughput example for a run with many PR lanes, not a required count)

- Gate A and gate B: Codex Sol medium, read-and-test only, working down `gate/queue.md`.
- Ship: Codex Sol medium, a 10-minute monitor that writes `SHIP-QUEUED`, `SHIP-FROZEN`, `SHIPPED`, `SHIP-STOP` lines.
- Reviewers: Astra xhigh for payments, privacy, sign-in and a rename with a migration; Sol high for routine reviews.

## A release sequence that worked (payments, sales closed; other releases follow their own runbook)

1. Plan → Astra review → back end (Codex) and UI (Opus) in test mode → gate PASS → Astra pre-merge CHANGES NEEDED (three findings: a recovery cursor, an account-delete guard, a lost first response) → fixes → gate PASS → Astra round 2 SAFE TO MERGE.
2. Release packet → first item on the morning list → the user's yes for "run the migration first, then merge".
3. The migration alone: dry run listed only that migration, applied, read back (history, row-level security, grants, foreign keys, functions, triggers).
4. The exclusive ship stopped on 27 failures the gate's scope had missed → lane fix → gate with every affected suite + Astra round 3 on the delta → exclusive ship, remote readback.

## Kinds of grants the run had

Each was dated, quoted in the run record and limited to that run: ship ordinary gated batches; a fast fix loop for error-tracker incidents, including rollback to a version known to work with the current database; urgent fixes merged after the lane's own checks; the migrations named in one question; one night's delegation with a 40% quota stop; an image budget for one design round; one payments release. None of them is a default.

## Files drift

Early rule files kept older instructions (a symlinked `node_modules`, 3 heavy slots) after later rules replaced them. Newer rules win; say which file you followed in the run record.
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: 472kUpdated Claude CodeCodexVerifiedInstall
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: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
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 availableStars: 322kUpdated Claude CodeCodexVerifiedInstall