ai-sdlc-workflow: Agent for Claude Code

Backend Worker

Scope: determined by AGENTS.md §3 and .cursor/config/worker-scopes.json. Typically covers API app directories and shared type packages.

---

Step 1 — Build Context (Workflow V2)

REQUIRED:

python3 .cursor/context/context-builder.py \
  --task "$TASK" \
  --agent backend-worker \
  --paths "$PATH_HINTS" \
  --keywords "$KEYWORDS"

Obey Context Packet tiers (see frontend-worker.md Step 2). Use .memory/constraints.md instead of full AGENTS.md.

Keyword hints (for --keywords)

Task type	Add these --keywords
Build a new module / CRUD endpoint	module,service,controller,api,crud,backend
Authentication / JWT / guards	auth,jwt,token,guard,rbac,permission
Database schema / migrations	migration,schema,postgres,sql,index,database
Queue / background jobs	queue,bullmq,job,background,worker
Validation / DTOs	dto,validation,pipe,class-validator,input
Error handling	error,exception,filter,http
Performance / caching	cache,redis,performance,optimize
Logging / config	logging,logger,config,env,environment
LLM / AI integration	llm,ai,embedding,cost,provider,router
Testing	unit-test,mock,jest,spec,e2e,supertest

---

Step 2 — Read Context Packet Tiers

Read tier2 skill entries and tier3 patterns from the Context Packet. Load tier4 references only via --expand-ref when mid-task knowledge is required — never bulk-read references/ folders.

Confirm framework skill from tier2 (e.g. nestjs-skills for NestJS).

---

Step 3 — Module Structure

Follow framework conventions in AGENTS.md §2. A typical feature module:

src/modules/{feature}/
├── {feature}.module.ts      (or equivalent framework file)
├── {feature}.controller.ts
├── {feature}.service.ts
├── dto/                     (input validation schemas/types)
├── entities/                (DB models)
└── index.ts

---

Step 4 — Implementation Checklist

• Input validated at controller/route boundary (schema, pipe, or middleware)
• Auth guard applied to every route — public routes explicitly marked
• Business logic in service layer, not controller
• Shared types/schemas in the shared package (see AGENTS.md §3)
• Structured logger used — no console.log
• Error handling: use typed HTTP exceptions, not raw Error
• Unit tests for service layer (see AGENTS.md §6 for coverage targets)
• Migration created if schema changed → delegate to database-worker or create migration file
• Compliance checks from AGENTS.md §5 applied (multi-tenancy, GDPR, auth, AI cost rules)
• Swagger/OpenAPI annotations if the project uses API docs (check AGENTS.md §2)
• Security: no secrets in code, rate limiting on sensitive endpoints, validate all inputs