feature-research

# Feature Research ## Overview Use this skill to perform structured, source-grounded research before implementation. The goal is to reduce rework and clarify decisions for broad or high-impact features. ## When to use - The request is broad (architecture, cross-cutting, platform-level) - Multiple valid approaches exist - The user asks for "best practices", trade-offs, or recommendations - The team needs a decision record before implementation ## Research process 1. **Clarify scope** - What problem are we solving? - What constraints matter (time, compatibility, security, performance)? - What is explicitly out of scope? 2. **Map options** - Produce 2–4 realistic options (not one obvious answer) - Include a conservative baseline option 3. **Gather evidence** - Use authoritative sources first (official docs, standards, mature references) - Distinguish facts from assumptions - Flag unknowns explicitly 4. **Evaluate trade-offs** - Correctness and maintainability - Complexity and delivery risk - Security and performance implications - Migration and rollback impact 5. **Recommend a path** - Pick one option as primary recommendation - Explain why alternatives were not chosen - Define phased rollout (thin vertical slices) ## Output contract When delivering research, provide: - **Context:** problem statement, constraints, non-goals - **Options:** each option in 3–6 bullets - **Comparison:** pros/cons/risks table - **Recommendation:** chosen option and rationale - **Execution draft:** first increments and verification plan - **Open questions:** what must be decided before implementation ## Quality bar - No implementation without a clear recommendation or explicit uncertainty - No "best practice" claims without concrete evidence - Avoid one-sided analysis; include meaningful alternatives - Keep recommendations actionable: reference `npm run lint`/`npm run build` as verification steps, `@/*` imports, and the existing `src/` structure ## Verification checklist - [ ] Scope and constraints are explicit - [ ] At least 2 viable options are compared - [ ] Risks and rollback strategy are described - [ ] Recommendation is concrete and justified - [ ] Next implementation slices are proposed