Notre avis

Cette compétence orchestre un pipeline multi-IA qui guide l'utilisateur à travers un processus structuré de planification, revues séquentielles, implémentation et revues de code en utilisant différents modèles d'IA.

Points forts

Impose un workflow discipliné avec des étapes distinctes de planification, implémentation et revue.
Exploite plusieurs modèles d'IA (Sonnet, Codex) pour améliorer la qualité et détecter les problèmes tôt.
Automatise la gestion des états et les cycles de revue séquentiels pour garantir la rigueur.
Fournit une traçabilité claire via des fichiers d'état et des sorties à chaque étape.

Limites

Nécessite la présence de compétences spécifiques (`/review-sonnet`, `/review-codex`, `/implement-sonnet`) pour fonctionner pleinement.
Peut introduire une surcharge pour des tâches simples qui pourraient être réalisées plus rapidement sans revues multi-étapes.
Le pipeline rigide peut ne pas convenir à tous les workflows ou préférences d'équipe.

Quand l'utiliser

Utilisez cette compétence lorsque vous avez besoin d'un processus de revue fiable et multi-modèle pour des changements de code complexes exigeant une qualité élevée et une validation minutieuse.

Quand l'éviter

Évitez cette compétence pour des tâches triviales ou exploratoires où la rapidité prime sur une revue rigoureuse, ou lorsque les compétences sous-jacentes requises ne sont pas disponibles.

Exemples

New feature with multi-AI pipeline

I want to add a new user authentication feature using JWT. Run the multi-AI pipeline to plan, review, implement, and review it.

Refactor a module with quality reviews

Refactor the payment processing module to use a strategy pattern. Use the multi-AI pipeline to ensure thorough reviews and implementation.

Fix a bug with change validation

Fix the race condition in the order submission handler. Follow the multi-AI pipeline to plan and implement the fix with reviews.

name: multi-ai description: Start the multi-AI pipeline with a given request. Guides through plan -> review -> implement -> review workflow. allowed-tools: Read, Write, Edit, Bash, Glob, Grep

Multi-AI Pipeline Orchestrator

You are starting the multi-AI pipeline. Follow this process exactly.

Reference Documents

First, read the standards that guide all reviews:

skill/multi-ai/reference/standards.md - Coding standards and review criteria

Step 1: Clean Up Previous Task

Remove old .task/ directory if it exists:

rm -rf .task
mkdir -p .task

Step 2: Capture User Request

Write the user's request to .task/user-request.txt.

Step 3: Create Initial Plan

Write .task/plan.json:

{
  "id": "plan-YYYYMMDD-HHMMSS",
  "title": "Short descriptive title",
  "description": "What the user wants to achieve",
  "requirements": ["req1", "req2"],
  "created_at": "ISO8601",
  "created_by": "claude"
}

Step 4: Refine Plan

Research the codebase and create .task/plan-refined.json:

{
  "id": "plan-001",
  "title": "Feature title",
  "description": "What the user wants",
  "requirements": ["req1", "req2"],
  "technical_approach": "Detailed how-to",
  "files_to_modify": ["path/to/file.ts"],
  "files_to_create": ["path/to/new.ts"],
  "dependencies": [],
  "estimated_complexity": "low|medium|high",
  "potential_challenges": ["Challenge and mitigation"],
  "refined_by": "claude",
  "refined_at": "ISO8601"
}

Step 5: Sequential Plan Reviews

Run reviews in sequence. Fix issues after each before continuing:

Invoke /review-sonnet
- Read .task/review-sonnet.json result
- If needs_changes: fix issues in plan, update .task/plan-refined.json
Invoke /review-codex
- Read .task/review-codex.json result
- If needs_changes: fix issues and restart from step 5.1
- If approved: continue to implementation

Step 6: Implement

Invoke /implement-sonnet

This skill will:

Read the approved plan from .task/plan-refined.json
Implement the code
Add tests
Output to .task/impl-result.json

Step 7: Sequential Code Reviews

Run reviews in sequence. Fix issues after each before continuing:

Invoke /review-sonnet
- Read .task/review-sonnet.json result
- If needs_changes: fix code issues
Invoke /review-codex
- Read .task/review-codex.json result
- If needs_changes: fix issues and restart from step 7.1
- If approved: continue to completion

Step 8: Complete

Write .task/state.json:

{
  "state": "complete",
  "plan_id": "plan-001",
  "completed_at": "ISO8601"
}

Report success to the user with:

Summary of what was implemented
Files changed
Tests added

Important Rules

Follow this process exactly - no shortcuts
Fix ALL issues raised by reviewers before continuing
If codex rejects, restart the review cycle from sonnet
Keep the user informed of progress at each major step

State Files Reference

| File | Purpose | |------|---------| | .task/user-request.txt | Original user request | | .task/plan.json | Initial plan | | .task/plan-refined.json | Refined plan with technical details | | .task/impl-result.json | Implementation result | | .task/review-sonnet.json | Sonnet review output | | .task/review-codex.json | Codex review output | | .task/state.json | Pipeline state |

Reference Directory

| Path | Purpose | |------|---------| | skill/multi-ai/reference/standards.md | Review criteria and coding standards | | skill/multi-ai/reference/schemas/ | JSON schemas for structured output |

Orchestrateur de Pipeline Multi-IA

Recommandé pour