ai-sdlc-workflow: Command for Claude Code

Act as Orchestrator driving @architect-planner and @judge-agent. Stay inside the current .cursor workflow.

Mode selection

• GENESIS — /architecture-plan brainstorming {idea}
• BREAKDOWN — /architecture-plan (active plan exists)

---

GENESIS mode — /architecture-plan brainstorming {idea}

Input: {idea}

Step 0 — Build context

python3 .cursor/context/context-builder.py --phase brainstorm --task "{idea}" --agent architect-planner
python3 .cursor/context/context-builder.py --phase plan --task "{idea}" --agent architect-planner

Step 1 — BRAINSTORM (HARD-GATE, chat only)

Concept brief (chat):

• What it does — job-to-be-done
• Target users
• Core features — MVP vs Later
• Suggested profile — frontend | backend | fullstack | library | cli
• Suggested layers on/off — frontend, backend, database, multi-tenancy, queue, auth, devops, ai-llm
• Design/UX needs — only if frontend layer likely on
• Domain & data — only entities/tenancy if database or multi-tenancy likely on
• Constraints
• 2–3 architecture approaches + recommendation

⏸️ STOP — wait for explicit approval of concept + profile/layers + MVP boundary.

Step 2 — STANDARDIZE DOMAIN (fill AGENTS.md)

After approval:

1. Write §0 Project Profile first (profile, layout, layers table).
2. Fill §1–§3 from the approved concept.
3. Fill optional sections only when layer is on:
   • multi-tenancy on → §4; else N/A — layer multi-tenancy off
   • database on → §9 entities; else N/A
   • queue on → §14; else N/A
   • ai-llm on → AI rules in §6; else skip
4. §5 roster: list agents that match active layers (+ always-on planner/judge/qa).
5. Align config:

python3 .cursor/context/profile-sync.py --from-agents

Optionally tighten worker-scopes.json to §3 paths (orchestrator approval for expansion).

Step 3 — CREATE ARCHITECTURE + ADRs

• docs/architecture.md — only containers/flows for active layers
• ADRs for foundational decisions that apply (do not invent tenancy/DB ADRs if those layers are off)