| name | implement |
| description | Orchestrates a coordinated team of specialized agents to implement plans sequentially from a given plans directory. Handles partial, missing, and complete plan states. Enforces a mandatory multi-reviewer loop (Code Reviewer, Doc Sync Reviewer, Security Audit Reviewer) after each phase before advancing. Use when the user says "implement", "/implement", or asks to implement plans from a directory. |
| arguments | [{"name":"plans_path","description":"Path to the directory containing the implementation plan files (e.g. @docs/frontend/plans/ or @docs/backend/plans/)","required":true}] |
Create a coordinated team of specialized agents to implement the plans in $ARGUMENTS, starting with Plan A and proceeding sequentially.
- If a plan has partial missing features: implement ONLY the missing features, then proceed to the next plan.
- If a plan has no implementation (all features pending): implement it fully and STOP. Do not proceed to any subsequent plans.
- If a plan is fully implemented: skip it and proceed to the next plan.
Agent Model Requirements:
- Code agents and code review agents MUST use Claude Opus.
- All other reviewer agents (doc sync reviewers, security audit reviewers) MUST use Claude Sonnet.
Review Process (mandatory after each plan phase is implemented):
- Spawn the following reviewers after each phase: Code Reviewer, Doc Sync Reviewer, Security Audit Reviewer.
- Reviewers MUST loop until ALL reviewers unanimously approve before advancing to the next phase.
- Do NOT advance until full consensus is reached.
