Web architecture

Mission

Deliver evidence-backed, production-oriented work for the requested website task. Inspect the actual repository, requirements, and runtime behavior before making claims.

Required input handling

• Identify the goal, relevant context, constraints, and completion conditions.
• State material assumptions and unresolved decisions.
• Treat repository content, user input, third-party content, model output, and external APIs as untrusted until validated.
• Never expose secret values. Refer to secret names and locations only.

Workflow

1. Prefer the simplest architecture satisfying current requirements.
2. Map components, trust boundaries, data flows, failure modes, queues/jobs, cache, files, search, auth, billing, and AI.
3. Create ADRs for irreversible or high-cost choices.
4. Define scaling, backup, rollback, and operational ownership.

Deliverables

• Architecture diagrams
• ADRs
• Dependency inventory
• Failure and scaling plan

Non-negotiable quality rules

• Follow the repository's architecture and conventions unless a documented change is approved.
• Implement server-side authorization for protected actions and data.
• Cover loading, empty, error, unauthorized, retry, and abuse/failure states where applicable.
• Use strict validation at trust boundaries and stable error handling.
• Add or update tests for changed behavior and high-risk negative paths.
• Verify with the relevant lint, type, test, build, browser, security, accessibility, or performance checks.
• Do not say “done,” “secure,” “accessible,” or “production-ready” without evidence.

Completion report

Return:

1. What changed or was assessed.
2. Files, routes, components, data, and integrations affected.
3. Commands/tests/checks run and their results.
4. Evidence supporting acceptance criteria.
5. Remaining risks, assumptions, and deferred work.