agentleFS
AsiaOstrichAsiaOstrich/EngramGraph/.agents/skills/orchestrate/SKILL.md

orchestrate

Skill4 starsChanged 29 days agoMITView on GitHub

Orchestrate multi-task execution plans using Claude's native Agent tool (DAG-based, no external engine). Use when: executing a plan.json file with parallel/sequential task dependencies. Not for: producing the plan in the first place — use /plan; single tasks with no dependencies between them. Keywords: orchestrate, plan, execute, DAG, task plan.

  • Deletes or force-pushes
  • Commits and pushes

161 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use define-requirements, orchestrate-plan and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • sson0-erSkill

    define-requirements

    Work with the user, one question at a time, to complete the requirements document (docs/claudex/<slug>/01-requirements.md). Run by the orchestrator (parent). Stays out of design and implementation.

    Also in 3 projects that use this

  • EternallLightSkill

    orchestrate-plan

    Use when executing a locked, reviewed PLAN.md by delegating the implementation to visible worker agents in Orca, herdr, or cmux ONE PHASE at a time — splitting the plan into context-window-sized phases, giving every phase a pristine worker context, and (when the plan touches INDEPENDENT repos) running one worker per repo IN PARALLEL. Triggers on "run the plan with a worker", "execute the plan in the other pane", "split the plan into phases and build it with workers", "build the backend and frontend in parallel panes". Not for edits you can finish inline; outside Orca, herdr, or cmux use tgc-skills:orchestrate.

    Also in 3 projects that use this

  • sson0-erSkill

    record-evidence

    How to record external facts (docs, command output, codebase findings) as evidence and design choices as decisions. Used by the researcher, designer and reviewer agents.

    Also in 3 projects that use this

  • PinkneyJiangSkill

    improve-architecture

    扫描代码库中的架构摩擦点,识别浅模块并挖掘深模块演进机会。生成可视化交互式 HTML 诊断报告,并在用户确认后闭环实施深度重构。

    Also in 4 projects that use this

  • kjnezSkill

    ticket

    Work on a JIRA/Linear ticket end-to-end: read ticket details, explore codebase, create a branch, implement with TDD, run quality checks, update ticket status, and create a PR. Use when the user provides a ticket ID to implement. Triggers on requests like "/ticket PROJ-123", "work on this ticket", "implement PROJ-123".

    Also in 4 projects that use this

  • xfstudioSkill

    reverse-engineer

    Expert reverse engineer specializing in binary analysis,

    Also in 3 projects that use this

The file

Files
SKILL.md

/orchestrate — 任務計畫編排 Skill

用法

/orchestrate <plan.json>           # 執行計畫
/orchestrate <plan.json> --dry-run # 僅驗證,不執行

執行流程

Step 1: 讀取與解析計畫(Claude-native)

使用 Read tool 讀取 plan.json,直接由 Claude 解析:

  1. 讀取 JSON 檔案,解析為任務清單
  2. 驗證結構:tasks 非空、每個 task 含 id、title、spec
  3. 安全掃描:檢查 task.spec 是否含危險指令(rm -rf /、DROP DATABASE、git push --force)
    • 若發現安全問題,列出所有問題並詢問使用者是否繼續
  4. 拓撲排序:依 depends_on 欄位將 tasks 分組為執行層(layers)
    • Layer 0:沒有依賴的 tasks(可並行)
    • Layer N:依賴全部在 Layer 0..N-1 中的 tasks

若 validation.valid 為 false,輸出錯誤訊息並停止。


Step 2: 呈現執行計畫

向使用者顯示解析結果:

計畫:<project>  品質:<quality>
──────────────────────────────────
Layer 0 (並行):T-001, T-002
Layer 1 (序列):T-003 (依賴 T-001)
Layer 2 (序列):T-004 (依賴 T-002, T-003)
──────────────────────────────────
共 4 個任務,3 個執行層

若使用 --dry-run:顯示以上摘要後立即停止,不啟動任何 Agent。


Step 3: 逐層執行

對 layers 陣列中的每一層:

  1. 同層並行:同一層的 tasks 在同一訊息中啟動多個 Agent tool
  2. Agent tool 參數:
    • prompt:task.spec + acceptance_criteria + user_intent 組合而成的執行提示
    • isolation: "worktree":每個 task 在獨立 git worktree 執行
    • description:"Task {task.id}: {task.title}"
  3. 等待完成:所有同層 agents 完成後,才進入下一層

Agent 提示範本:

Task: {task.title}

Spec:
{task.spec}

Acceptance Criteria:
{task.acceptance_criteria 逐條列出}

User Intent:
{task.user_intent}

After completing the task, run: {task.verify_command}

Step 4: 依賴失敗處理

如果某個 task 的 agent 回報失敗:

  • 標記該 task 為 failed
  • 後續層中 depends_on 包含該 task ID 的所有 tasks 標記為 skipped
  • 繼續執行不受影響的 tasks
  • 最終報告中標示失敗原因

Step 5: Quality Gate

每個 task 完成後,若 task 定義了 verify_command:

  1. 使用 Bash tool 執行 verify_command
  2. 成功:記錄 status: "success"
  3. 失敗:
    • 若 quality 為 "strict" 或 "standard":啟動 Fix Loop
    • Fix Loop:將錯誤輸出回傳給 agent,重新執行(最多 3 次)
    • 超過重試上限:標記為 failed

Step 6: Judge 審查(選用)

如果 task 定義了 judge: true:

  • 啟動一個額外的 Agent tool 進行審查
  • 提示:審查 task {id} 的 git diff,依照 code-review 標準檢查
  • Judge 結果附加到最終報告

Step 7: 產出報告

彙整所有 task 結果,輸出執行報告:

{
  "summary": {
    "total_tasks": 4,
    "succeeded": 3,
    "failed": 1,
    "skipped": 0,
    "total_duration_ms": 120000
  },
  "tasks": [
    { "task_id": "T-001", "status": "success", "duration_ms": 30000 },
    { "task_id": "T-002", "status": "failed",  "error": "..." },
    { "task_id": "T-003", "status": "skipped", "reason": "依賴任務失敗" }
  ]
}

品質設定檔

Profile行為
strict全部檢查 + Fix Loop 最多 5 次
standardlint + test + Fix Loop 最多 3 次(預設)
minimal僅跑 verify_command,不 Fix Loop
none不執行任何驗證

重要注意事項

  • 每個 Agent tool 的 isolation: "worktree" 會自動建立 git worktree
  • depends_on: [] 的 tasks 在同一層內全部並行
  • 使用繁體中文回覆使用者

版本歷程

版本日期變更
2.0.02026-04-28升格為 UDS 正式 Skill;Step 1 改為 Claude-native 解析(移除 plan-resolver-cli.js 依賴);新增 Quality Gate / Fix Loop 章節(XSPEC-097 Phase 3)
1.0.02026-04-09初始發布(從上游遷移)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.