agents reference

Agents Reference

� Navigation : README → Agents Reference

Catalogue complet des agents AI Agent Kit avec leurs spécialités et cas d'usage

#tags: agents, reference, catalog
#difficulty: intermédiaire
#time: 15-30 minutes
#category: reference-technique | @python-developer | backend/python/python-developer.md | Python development | FastAPI, Django, Clean Architecture | | @java-developer | backend/java/java-developer.md | Java development | Spring Boot, Maven | | @go-developer | backend/go/go-developer.md | Go development | Clean Architecture, Gin, Echo | | @rust-developer | backend/rust/rust-developer.md | Rust development | Tokio, Axum, Cargo | | @react-developer | frontend/react/react-developer.md | React development | TypeScript, TanStack Query | | @vue-developer | frontend/vue/vue-developer.md | Vue.js development | Composition API, Pinia | | @svelte-developer | frontend/svelte/svelte-developer.md | Svelte development | Svelte 5, SvelteKit | | @nextjs-developer | frontend/nextjs/nextjs-developer.md | Next.js development | App Router, Server Components | | @nuxt-developer | frontend/nuxt/nuxt-developer.md | Nuxt development | Vue 3, File-based Routing | | @semantic-searcher | agents/semantic-searcher.md | Recherche sémantique | @semantic-searcher "concept recherché" | | @code-explorer | agents/code-explorer.md | Analyse architecturale | @code-explorer "architecture projet" |

🚀 Usage Examples

Universal Format (All Tools)

# Copy from .ai-agents/agents/common/architect.md
@architect
Design payment processing module with Clean Architecture

Claude Code Format (Optimized)

/agents                    # List available agents
@architect "Design payment processing module"

📖 Agent Details

Rôle : Design l'architecture et définit les contrats.

@architect "Design payment processing module"
@architect "Design Clean Architecture for Order management"

Output : Design document avec layers, contracts, next steps.

Gates : SOLID compliance, Dependency Rule.

---

Developer

Rôle : Implémente le code de production.

@developer "Implement User entity with email validation"
@developer "Implement Order service with payment integration"

Output : Fichiers créés, build status.

Gates : dotnet build --warnaserror, dotnet format.

Personnalisation :

agents:
  developer:
    instructions_append:
      - "Use primary constructors for DI"
      - "Use file-scoped namespaces"

---

Tester

Rôle : Écrit des tests automatisés.

@tester "Write tests for User entity"
@tester "Write integration tests for Products API"

Output : Fichiers de tests, coverage report.

Gates : dotnet test, coverage >= 80%.

Personnalisation :

agents:
  tester:
    instructions_append:
      - "Use Bogus for test data"
      - "Achieve 90%+ coverage"

---

Reviewer

Rôle : Revue de code complète.

@reviewer "Review Order implementation"
@reviewer "Check security of AuthController"

Output : Review report avec décision (APPROVE / REQUEST_CHANGES / REJECT).

Gates : All gates pass, security check.

Décisions :

• APPROVE → Continue to docwriter
• REQUEST_CHANGES → Return to developer
• REJECT → Return to architect

---

Docwriter

Rôle : Met à jour la documentation.

@docwriter "Update documentation for Order feature"
@docwriter "Document new Products API endpoints"

Output : Fichiers mis à jour, impact assessment.

Impact levels :

• HIGH_IMPACT : New feature, breaking change → Full update
• MEDIUM_IMPACT : Config change → Targeted update
• LOW_IMPACT : Internal refactor → Verify only

---

Token Optimizer

Rôle : Analyse la tâche et optimise l'exécution des agents.

# Appelé automatiquement par @chain:optimized-feature
# Ou manuellement pour analyser une tâche
@token-optimizer "Add validation to User entity"

Output : Plan optimisé avec agents à utiliser/skipper.

Analyse :

• Détecte le type de tâche (quickfix, bugfix, feature, refactor, docs, test)
• Identifie les layers impactés (Domain, Application, Api, Infrastructure)
• Skip les agents non nécessaires
• Charge seulement les sections de contexte pertinentes

Skip conditions :

Agent	Skip si...
architect	Changement localisé, pas nouvelle entité
unit-tester	Config/docs uniquement
integration-tester	Pas de changement API
docwriter	Refactor interne
reviewer	Changement < 5 lignes

---

Agents spécialisés

Agent	Usage	Tokens
@token-optimizer	Optimisation tokens, skip agents inutiles	~30
@explorer	Exploration read-only du codebase	~40
@unit-tester	Tests unitaires uniquement	~70
@integration-tester	Tests d'intégration uniquement	~70

---

Configuration globale

agents:
  <agent_id>:
    model: "claude-4-5-opus-20241022"
    max_tokens: 8192
    temperature: 0.3
    instructions_append:
      - "Additional instruction"
    gates_override:
      build:
        required: true

→ Voir Chains Reference pour les workflows