agentleFS
Sign inStart free
juandelossantosjuandelossantos/another-agent-skills/skills/multi-agent-orchestration/SKILL.md

multi-agent-orchestration

Skill1 starsChanged 11 days agoMITView on GitHub

Orchestrate multiple AI coding agents in parallel, pipeline, or swarm modes. Use with >2 agents or multi-file refactors. Do NOT use for single-agent tasks.

We split this skill into the rules it gives an agent

3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 14 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Meta-skill for parallel and multi-phase workflows: Not a platform skill — augments the Orchestrator (primary session agent).

    line 17
  2. Very specific:

    If parallelizing: reference target skill per subagent (frontend-web, backend-api-mastery, etc.)

    line 56
  3. Distinctive:

    No error handling when subagent fails

    line 76

“Common” is measured: kept by at least 6 in 10 of the 14 edited versions of this file we could read. “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

Multi-Agent Orchestration

Meta-skill for parallel and multi-phase workflows. Not a platform skill — augments the Orchestrator (primary session agent). Does NOT replace engineering-fundamentals or any platform skill.

Output Contract

Merged, tested source code from coordinated subagents. No single artifact — the orchestration process produces the final integrated result.

Note: This meta-skill has 1 guide (GUIDE.md) by design — it documents orchestration patterns, not phased workflows. Platform skills need ≥ 2 guides; meta-skills don't follow that pattern.

When to Use

Invoke when:

  1. Task requires > 2 agents (parallel implementation, research+build+review)
  2. Multi-module refactor (non-overlapping files, independent modules)
  3. Spec-driven pipeline with phase gates (Spec → Implement → Test → Review)
  4. Competitive exploration (swarm mode for architecture decisions)

When NOT to Use

  • Single-file changes (one agent is sufficient)
  • Tasks where files overlap (causes merge conflicts)
  • Simple CRUD or linear workflows

Quick Workflow

  1. Detect — Is this task parallelizable? Check overlap. If files overlap → sequential.
  2. Decompose — Break into non-overlapping sub-tasks. Plan context per subagent.
  3. Delegate — Launch subagents via task tool. Pass only relevant skill + context.
  4. Collect — Receive results. Verify each subagent's output independently.
  5. Integrate — Merge files. Resolve conflicts. Run build/tests.
  6. Verify — Full integration test. Only Orchestrator commits.

Load Order

  1. This SKILL.md (index)
  2. guides/GUIDE.md — Full patterns, real examples, error recovery
  3. If parallelizing: reference target skill per subagent (frontend-web, backend-api-mastery, etc.)

Relation to Other Skills

This skillOther skillWhy
Orchestrator delegates buildincremental-implementationSubagent 1 builds module A
Orchestrator delegates teststest-driven-developmentSubagent 2 writes tests
Orchestrator delegates reviewcode-review-and-qualitySubagent 3 reviews output

Red Flags

  • Subagents editing the same file simultaneously
  • Subagents receiving full AGENTS.md
  • Orchestrator commits without verifying subagent output
  • explore agent type used for writing code (read-only!)
  • No error handling when subagent fails

Verification

  • Sub-tasks have non-overlapping file assignments
  • Each subagent receives only its relevant skill
  • Subagent output verified before integration
  • Orchestrator merges, tests, then commits (never subagents)
  • Build + tests pass after integration

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.

Reviews can't be read right now.

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.

Check your setup, free