plan-interview — Claude Code skill: interview to spec
Turns a rough plan into an implementation-ready spec through an adaptive interview. Useful when the idea exists but the requirements don't.
npx skills add secondsky/claude-skills --skill plan-interview -a claude-codeWhat plan-interview does
The skill runs a structured interview over your plan and writes a spec file next to it. Depth scales with complexity: 10-15 questions for a single clear feature, 18-23 when several components and integrations are involved, 22-28 for cross-system work with many stakeholders. Questions move through phases — scope and success criteria first, then architecture and data, then user flows, risks and operations.
You start it with the slash command /plan-interview:interview [plan-file]. It reads your plan file and produces something like docs/feature-plan-spec.md. A spec-reviewer subagent scores the finished spec across four dimensions — completeness, consistency, clarity and edge cases — 25 points each. Ask it to review your spec and it runs.
The interview skips whatever the plan already answers. Instead it probes edge cases, surfaces assumptions you never wrote down, and flags answers that contradict each other. If the plan is barely a paragraph, expect a long session — that's a sign to fill in scope and goals first, or half the questions will just restate the obvious.
What it helps with
- turn a feature sketch into a spec you can code against
- gather requirements through ordered questions instead of a messy list
- surface hidden assumptions and contradictions before implementation starts
- score a finished spec for completeness and edge cases
- pin down risks and rollback plans before the first commit
Example prompts for the agent
- /plan-interview:interview docs/feature-plan.md
- interview me about the plan in docs/payments-plan.md and save the spec next to it
- review my spec for contradictions and missing edge cases
How to install plan-interview
1Skills CLI (Vercel)
npx skills add secondsky/claude-skills --skill plan-interview -a claude-code2Manually: put the files in place
~/.claude/skills/plan-interview/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/plan-interview/
1Skills CLI (Vercel)
npx skills add secondsky/claude-skills --skill plan-interview -a codex2Manually: put the files in place
.agents/skills/plan-interview/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add secondsky/claude-skills --skill plan-interview -a cursor2Manually: put the files in place
.cursor/skills/plan-interview/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install secondsky/claude-skills plan-interview2Manually: put the files in place
.github/skills/plan-interview/SKILL.md1Skills CLI (Vercel)
npx skills add secondsky/claude-skills --skill plan-interview -a gemini-cli2Platform CLI
gemini skills install https://github.com/secondsky/claude-skills --path plugins/plan-interview/skills/plan-interviewManually: put the files in place
.devin/skills/plan-interview/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add secondsky/claude-skills --skill plan-interview -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about plan-interview
How do I install plan-interview in Claude?
Run npx skills add secondsky/claude-skills --skill plan-interview -a claude-code. The skill file lands at ~/.claude/skills/plan-interview/SKILL.md.
What does plan-interview do?
It interviews you about your plan and assembles the answers into an implementation-ready specification. Question depth follows the complexity of the plan.
How does the spec-reviewer subagent work?
It scores a finished spec on four dimensions — completeness, consistency, clarity and edge cases — 25 points each. It triggers when you ask to check spec quality.
Source text
SKILL.md · MIT
--- name: plan-interview description: Adaptive interview-driven spec generation. Use when converting rough plans into comprehensive specifications, needing structured requirements gathering, or transforming ideas into implementation-ready documentation. license: MIT --- # Plan Interview Skill Transform rough plans into comprehensive, implementation-ready specifications through adaptive, structured interviews. ## When to Use - Converting a plan or idea into a detailed specification - Gathering requirements through structured questioning - Transforming rough documentation into implementation-ready specs - Ensuring all edge cases, risks, and stakeholders are considered before implementation ## Available Components ### Command: `/plan-interview:interview [plan-file]` Adaptive interview that calibrates depth based on plan complexity: | Complexity | Signals | Questions | |------------|---------|-----------| | **Simple** | Single feature, clear scope | 10-15 | | **Moderate** | Multi-component, some integrations | 18-23 | | **Complex** | Cross-system, many stakeholders | 22-28 | **Usage:** ```bash /plan-interview:interview docs/feature-plan.md # Output: docs/feature-plan-spec.md ``` ### Agent: `spec-reviewer` Autonomous quality analysis of specifications across 4 dimensions: - **Completeness** (25 pts) - All sections populated? - **Consistency** (25 pts) - No contradictions? - **Clarity** (25 pts) - No ambiguous language? - **Edge Cases** (25 pts) - Error handling defined? Triggers when you say "review my spec" or "check specification quality". ## Interview Phases 1. **Foundations & Scope** - Stakeholders, success criteria, constraints, MVP scope 2. **Technical Deep-Dive** - Architecture, data models, scalability, security 3. **User Experience** - Personas, flows, cognitive load, error recovery 4. **Risks & Tradeoffs** - Risk categorization, blast radius, contingency plans 5. **Operationalization** - Testing, deployment, monitoring 6. **Wrap-Up** (optional) - Only for complex plans with remaining gaps ## Interview Philosophy **Core Principle**: Depth over breadth. Better to deeply understand critical aspects than superficially cover everything. **Key Techniques**: - Non-obvious questions - Skip what the plan already answers - Edge probing - What happens in unusual cases? - Assumption surfacing - Make implicit beliefs explicit - Contradiction detection - Flag when answers don't align - Adaptive depth - Probe deeper on complex areas, move faster on clear ones ## Spec Output Structure Generated specs include: - Overview (problem, solution, success criteria, stakeholders) - Functional and non-functional requirements - Technical design (architecture, data models, APIs, security) - User experience (personas, flows, states, edge cases) - Risks and mitigations (risk register, tradeoffs, contingency plans) - Implementation notes (key decisions, dependencies, migration) - Operationalization (testing, deployment, monitoring) - Open questions and out-of-scope items - Phasing (MVP vs future) ## References Load these for deeper guidance during interviews: - `references/phase-1-clarifications.md` - Foundations questions and pitfalls - `references/phase-2-technical.md` - Architecture discussion patterns - `references/phase-3-ux.md` - Persona development, UX patterns - `references/phase-4-risks.md` - Risk assessment frameworks - `references/interview-techniques.md` - Cross-cutting interview skills - `references/example-spec.md` - Annotated high-quality spec example
© secondsky · MIT · secondsky/claude-skills
