designlisted

# Step 2: Design (วางโครงสร้าง) **Habit**: H8 — Find Your Voice | **Anti-pattern**: Letting AI decide architecture without human judgment ## Process 1. **Read existing architecture and context contract**: Check `CLAUDE.md`, `AGENTS.md`, `SPEC.md`, `DOMAIN.md`, `CONTEXT.md`, `CONTEXT-MAP.md`, `DESIGN.md`, `ARCHITECTURE.md`, `docs/agents/domain.md`, and ADR directories. Understand current state and project vocabulary before proposing changes. 1b. **Validate scope alignment**: If a `SKILL_OUTPUT:requirements` block exists in the PRD output, read it and verify: - Proposed architecture decisions don't expand beyond `scope_in` - Success criteria are achievable with the proposed design - Identified `risks` are addressed or accepted in design constraints 1c. **Select the smallest safe pass level** before producing design output: | Pass level | Use when | Expected output | | ---------- | -------- | --------------- | | **Scan** | Small, bounded, exploratory, or unclear architecture impact | Compact architecture note, key constraints, open questions, and safe next step | | **Focus** | One module, workflow, subsystem, integration, or boundary is in scope | Targeted decisions, local trade-offs, risks, and any needed ADRs | | **Full** | Whole-system design, unclear ownership, 3+ interacting modules, persistence, authentication, payment, security, deployment, or future-agent handoff is needed | Full decision set, ADR coverage, risk register, human approvals, and handoff-ready constra