agentleFS
rumotionrumotion/ai-agent-project-template/.agents/skills/delegation-coordinator/SKILL.md

delegation-coordinator

Skill1 starsChanged 43 days agoMITView on GitHub

Coordinates task delegation across Gemini, Codex, Claude, and subagents using the portable subagent contract.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Keep one lead and one worker by default, at most two open tasks unless the plan explicitly increases that limit.

    line 13
  2. Very specific:

    Run parent verification checks and diff the worktree against owned_paths before marking delegated tasks completed.

    line 20
  3. Distinctive:

    Single Writer Rule: Ensure no two concurrent subagents own the same target file path for editing.

    line 24
  4. Very specific:

    Deterministic Verification: Verify subagent results with explicit build, test, or lint commands before accepting the result.

    line 27
  5. Very specific:

    Memory Bank Serialization: Only the parent writes memory-bank/*.

    line 28
  6. Very specific:

    Conflict Handling: If a worker returns status: blocked or fails verification twice, halt and escalate to parent/Codex review.

    line 31

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Delegation Coordinator

Coordinates task breakdown, path ownership, worker dispatch, and evidence-grounded result validation across multi-agent sessions.

Quick Start

  1. Apply the completion-first admission rules in docs/subagent-contract.md. Keep one lead and one worker by default, at most two open tasks unless the plan explicitly increases that limit. Stop dispatch while review or blocked work is pending; workers do not spawn workers.
  2. Define worker tasks using the JSON task envelope validated by schemas/subagent-task.schema.json (rules in docs/subagent-contract.md).
  3. Assign disjoint owned_paths to each worker to prevent write collisions; list hard exclusions in forbidden_paths.
  4. Dispatch the envelope only — never forward the parent transcript, other tasks, or secret-bearing history.
  5. Require workers to return the result envelope validated by schemas/subagent-result.schema.json, capped by max_output_tokens.
  6. Run parent verification checks and diff the worktree against owned_paths before marking delegated tasks completed.

Delegation Rules

  • Single Writer Rule: Ensure no two concurrent subagents own the same target file path for editing.
  • Read-Only Workers: Set read_only: true for audit, critic, or research workers.
  • Envelope-Only Context: Workers receive only their JSON task envelope; mask secrets before dispatch.
  • Deterministic Verification: Verify subagent results with explicit build, test, or lint commands before accepting the result; reject every write outside owned_paths or within forbidden_paths, including reported writes. Compare content hashes for already-dirty files, not just status.
  • Memory Bank Serialization: Only the parent writes memory-bank/*; merge worker-reported facts once.
  • Completion Before Expansion: Dependencies release only after parent acceptance. Reserve time and every binding quota for implementation, parent checks and integration. With pacing, use a fresh completion board and dispatch only dispatch_task_ids; quota headroom alone never permits launch.
  • Directed Coordination: Workers send bounded progress, blockers and result references to the lead. The lead routes changed contracts and cancellation requests; no peer may change another worker's scope. A silent worker retains ownership until native execution is confirmed stopped.
  • Conflict Handling: If a worker returns status: blocked or fails verification twice, halt and escalate to parent/Codex review.

Validate paired results with scripts/validate-schemas.py --schema schemas/subagent-result.schema.json --instance result.json --task task.json --root .. Read-only tasks use owned_paths: []. Serialized-byte limits are independent of token telemetry; declarations do not create a sandbox or lease system.

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.