beads-coordinator
Coordinate unattended Beads issue execution — rank ready work finish-first (epics closest to done before new ones), claim atomically, and allocate cohesive delivery packages to persistent outcome owners, alone or as one of several coordinators sharing a backlog. Use for unattended issue throughput in Beads-backed repos.
- License 许可
- MIT
- 最近更新
- 2026-11-10
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- beads-coordinator
- metadata
- { "owner": "tze", "status": "active", "authors": [ "tze", "OpenAI Codex", "Claude Fable 5.1", "Claude Opus 5.5" ], "last_reviewed": "2026-11-06" }
- description
- Coordinate unattended Beads issue execution — rank ready work finish-first (epics closest to done before new ones), claim atomically, and allocate cohesive delivery packages to persistent outcome owners, alone or as one of several coordinators sharing a backlog. Use for unattended issue throughput in Beads-backed repos.
- compatibility
- Requires a Beads-backed repository with `bd` v1.3+ (`parent` on list rows, `bd epic status`, `--if-assignee`), `uv`, `git`, an authenticated `gh`, `jq`, git worktree support, and network access for `gh` PR operations.
02 资源文件 14
- 参考文档agents/openai.yaml320 B
- 参考文档README.md253 B
- 参考文档references/commands.md5.2 KB
- 参考文档references/coordinator-loop.md23.7 KB
- 参考文档references/epic-coordination.md2.1 KB
- 参考文档references/runtime-and-safety.md26.9 KB
- 脚本scripts/coordinator_registry.py11.4 KB
- 脚本scripts/normalize_pr_review_state.py43.4 KB
- 脚本scripts/rank_ready_work.py55.3 KB
- 脚本scripts/review_relation_graph.py2.7 KB
- 脚本tests/test_coordinator_registry.py7.2 KB
- 脚本tests/test_normalize_pr_review_state.py85.7 KB
- 脚本tests/test_rank_ready_work.py74.8 KB
- 脚本tests/test_workflow_contract.py18.4 KB
03 SKILL.md 正文
Beads Coordinator
Allocate bounded outcomes to persistent implementation owners and recover their delivery. This controller handles lifecycle writes for dispatched workers; implementation, diagnosis and routine corrections remain with the same owner. For a session doing its own delivery use beads-looper. For a supplied package use beads-worker.
Use for “keep N workers finishing the backlog”, “finish started outcomes first”, or “coordinate several owners on this tracker”. Backlog authoring routes to beads-writer, recovery-only to beads-cleanup.
Contract and reading order
- Shared delivery model: units, persistent owners, branch tradeoffs, packets, claim journal, evidence and acceptance.
- Coordinator loop: allocation, review wiring, corrections, landing, per-member reconciliation and reports. Start here after the shared model; load other references only for the step that needs them.
- Runtime and safety: actor/lease, capacity, guarded adoption, model selection and bounded waits.
- Epic coordination: shared prerequisites and finite programme ownership; no extra mini-coordinator tier.
- Commands: CLI lookup and session completion.
Claim with atomic --claim, use a unique actor and never mutate another live
actor's claims. Multi-bead packages use the shared helper and journal. Dispatch
native runtime sessions with absolute skill paths and concise current packet
locators. Require worker bootstrap proof before editing. Never commit Dolt-managed
.beads/ files, run production actions without authority, or bypass a repo gate.
Executable consumers
- rank_ready_work.py ranks finite outcomes and
prerequisites and review-first admission. Pass the Step 0 normalization file;
queueis executable order,candidatesdiagnostic only. Missing review evidence dispatches nothing; parked owners keep claims without consuming execution slots. - coordinator_registry.py is the registry writer and shared liveness reader; do not hand-write registry expiries.
- normalize_pr_review_state.py reports canonical anchor/review associations and terminal-state recommendations.
- review_relation_graph.py validates relations without I/O; partial relation/inventory evidence permits no mutation.
Target trackers using bd -C /absolute/repo on every command when outside that
repo. A bead prefix does not route across databases. Search
known errors for unexpected errors, append
confirmed new fixes there, and apply decision autonomy
when a blocker is actually an engineering choice. Use
token efficiency for projected output
and reuse of evidence rather than repeated full scans.
04 同仓库更多 Skills
user-run
Use when a step needs privileges this session lacks (docker, sudo, another user's checkout or secrets, host access) and the owner must run it as tze. Writes a reviewable /home/orca/.tmp/<slug>/run.sh that impersonates tze (USER_TO_IMPERSONATE overrides), tees timestamped logs plus latest.log and collects artifacts there; Claude presents it with a risk and time brief, then continues from the logs. Triggers: "I'll run it for you", "permission denied on docker.sock", "needs sudo", "sudo: a password is required", "run as tze", "/user-run".
requesting-code-review
Use when completing tasks, implementing major features, or before merging to verify work meets requirements
receiving-code-review
Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation
finishing-a-development-branch
Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup
executing-plans
Use when you have a written implementation plan to execute in a separate session with review checkpoints
dispatching-parallel-agents
Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies