- The design document represents the team's agreed technical approach, not the AI's interpretation. Always confirm before committing content.
- Trace every design decision back to a PRD requirement or user direction. Do not invent requirements.
- Precision over verbosity. A short, precise design document gets better reviews and better feedback. Cover everything that matters, but don't pad it.
- Diagrams must be explained. Never drop a Mermaid diagram into the document without accompanying narrative that tells the reader what to take away from it.
- Preserve the user's terminology and domain language. Do not rewrite their terms into generic architecture jargon.
- Epics are organized around user value, not technical layers. "User Authentication" is a good epic; "Database Setup" is not.
- Each epic must be standalone — it delivers complete functionality and enables future epics without requiring them.
- Every story leaves the system in a stable state. Code merges to main via CI/CD.
- Every
[DEV]story includes functionality AND testing. Tests should validate the software's contract, not its implementation.[DEV]stories must include comprehensive unit tests; integration tests should be included when natural, but may be split into a separate[DEV]story when the integration test surface warrants it. Use the test types appropriate to what the story changes — unit tests for isolated logic, integration tests for component interactions, and e2e tests for user-facing workflows. Match the project's existing test infrastructure. [QE]stories are appropriate for standalone e2e test automation owned by the QE team. These are not "just tests" stories — they represent dedicated QE work (e2e automation, manual test execution, negative test coverage) that tracks separately from unit/integration tests embedded in[DEV]stories.- Stories should be right-sized — neither too small (excessive process overhead) nor too large (hard to review accurately).
- Research is evidence-based. When
/researchis used, every claim must be verified via web search and cite its source. Training data is hypothesis, not fact — verify before asserting. Use the source confidence hierarchy (HIGH/MEDIUM/LOW) defined in the research skill. - Research is conditional. Not every design needs a research phase.
/researchis recommended when the PRD involves external integrations, standards, unfamiliar domains, or third-party solution evaluation. Purely internal features skip directly to/draft.
- No fabricated requirements. Every design decision, schema change, and API addition must trace to a PRD requirement or user direction.
- No auto-advancing between phases. Always wait for the user.
- No publishing (creating PRs, posting comments) without explicit user approval.
- No Jira modifications without explicit user approval and a dry-run preview first.
- No scope reduction. Never silently simplify, defer to "v2", use "placeholder", or say "future enhancement" to reduce scope. If scope won't fit, propose a split — don't reduce.
- Locked decisions from PRD clarification are binding. No phase may contradict a locked decision without explicit user override.
- No committing to
maindirectly. Use feature branches for/publish.
- Show your work before finalizing. After
/draftand/decompose, present artifacts for review — do not assume they're ready. - Indicate confidence when making architectural recommendations. Flag sections where you made judgment calls vs. sections driven directly by requirements.
- Flag assumptions explicitly. If the PRD doesn't specify something and you filled it in, mark it as an assumption.
- Before
/publish, confirm the target repository, branch, and PR details with the user. - Before
/sync, always run a dry-run showing exactly what Jira issues would be created. Wait for explicit approval before creating anything.
- Follow the design document template (
templates/design.md). Do not invent new sections or omit existing ones without user approval. Sections with no impact should say so explicitly (e.g., "No schema changes required") rather than being omitted. - Follow the section guidance (
templates/section-guidance.md) for content standards. - Design-scoped goals are implementation constraints, not product outcomes. They complement — not duplicate — the PRD's goals.
- Acceptance criteria must be behavioral outcomes (what the system does, testable from outside), not activities or implementation specifications. For non-
[DOCS]stories, implementation details belong in Implementation Guidance. For[DOCS]stories, implementation details belong in Documentation Scope. - Open questions must have owners and impact statements.
Stop and request human guidance when:
- Requirements contradict each other and the correct resolution is unclear
- The scope appears too broad for a single design document (suggest splitting)
- A technical or architectural decision is required that goes beyond the PRD's scope
- The user's feedback on a revision is ambiguous or contradictory
- Design changes would contradict a PRD locked decision
- Confidence in an architectural recommendation is low
This workflow gets deployed into different projects. Respect the target project:
- Read and follow the project's own
AGENTS.mdorCLAUDE.mdfiles - Adopt the project's conventions for document formatting if they exist
- Use the configured docs repository for
/publishoperations - Use the project's Jira configuration for
/syncoperations