What does this skill do?

The Multi-Agent RFC Skill Pipeline orchestrates the execution of RFCs that are too large for a single step by breaking them down into independent, verifiable work units. Each unit is assigned to an agent, follows a complete quality pipeline, and only enters the merge queue once it reaches the “merge-ready” state. The acyclic directed graph ensures a deterministic execution order and full traceability from the RFC to final integration.

Decomposition into a DAG
Divide an RFC into work units with explicit dependencies, forming an acyclic graph with a deterministic order.
Quality Pipeline by Unit
Conduct research, planning, implementation, testing, review, and preparation of the merge report for each unit.
Merge queue with validation
Merge only approved units into the integration branch, rerunning tests after each merge.
Recovery from Lockups
Remove stalled units from the queue, take a snapshot, and retry with a narrower range.

Usage examples

🔀 Break down the RFC
Analyze this authentication refactoring RFC and break it down into verifiable work units using its dependency DAG.
🧪 Run unit
Run the entire quality pipeline for the auth-token-rotation unit and generate the merge-ready report.
🛡️ Reset Lock
The payment integration unit has stalled; remove it from the queue and try again with a narrower scope.
📊 Verify DAG
Check the status of the entire DAG and provide me with the integration risk summary, including any unresolved dependencies.

Features

Directed Acyclic Graph (DAG) Define explicit dependencies between units with a deterministic execution order and no cycles.
Quality doors per unit Each unit goes through the following stages: research, planning, implementation, testing, review, and a "merge-ready" report.
Controlled merge queue Only units with a "merge-ready" status are queued; tests are rerun after each merge.
Automatic Recovery Capture snapshots of locked units and retry execution with a reduced scope and updated constraints.
Complete traceability Generates an execution log, scorecards by unit, a graph snapshot, and an integration risk summary.

Frequently asked questions

It is a standalone portion of the RFC with explicitly defined dependencies, acceptance tests, risk level, and rollback plan.
Only when it reaches the "merge-ready" status after passing all the validations in the quality pipeline and having no broken dependencies.
It is removed from the active queue, a snapshot of the findings is captured, a narrower scope is regenerated, and execution is retried.
No. The skill works with Claude Code and Cowork without any additional external dependencies; all orchestration is handled internally.
Multi-Agent RFC Pipeline — RFC Orchestration with Claude AI

¿Prefieres escuchar el contenido? Genera la narración de audio con un clic.