multi-agent-review-gate
Run a structured review gate before merging multi-agent outputs. Use when multiple agents have produced work packets that need quality checks, conflict detection, and approval before integration.
SKILL.mdresearch/sphere-grid/skills/multi-agent-review-gate/SKILL.md
GitHubMulti Agent Review Gate
Overview
Enforces a quality and consistency gate after parallel agents complete their work packets, ensuring outputs are compatible, correct, and safe to merge.
Review Checklist
- Completeness — Every work packet meets its
done_criteria. - Conflict scan — No two packets modified the same file in incompatible ways.
- Interface check — Cross-packet APIs, types, and contracts are consistent.
- Test validation — All packet-level and integration tests pass.
- Style & standards — Code follows project conventions (linting, formatting, naming).
- Security scan — No new secrets, injection vectors, or unsafe patterns introduced.
Gate Decisions
| Decision | Meaning |
|---|---|
| APPROVE | All checks pass — safe to integrate |
| REQUEST_CHANGES | Issues found — return to packet owner with specific feedback |
| BLOCK | Critical conflict or failure — escalate before proceeding |
Conflict Resolution
- If two packets touch the same file, diff both changes and determine if they're additive or conflicting.
- Additive changes can be merged in dependency order.
- Conflicting changes require one packet owner to rebase on the other's output.
When to Use
- After a multi-agent-orchestrator run completes all packets
- Before committing or pushing multi-agent work
- When integrating outputs from agents that worked in parallel on related code