Generate interview.md via Interactive Interview

Conduct a structured interview with the user to gather all context needed for a story, then generate the interview.md file that feeds into the story spec step (dispatched by /continue), /mockups, /api-spec, /test-spec.

Usage

/interview 5                    # By MVP story number
/interview "Create task"             # By story name
/interview                      # Interactive selection

Workflow

Phase 1: Context Gathering (Silent)

Before asking any questions, silently read. Resolve the target story first (Phase 2's parse step) — items 4, 6, 9 and 10 are scoped to it and cannot be read before it is known:

1. Story mapping: ProductSpecification/stories.md
2. Product description: ProductSpecification/BriefProductDescription.md
3. Expected load: ProductSpecification/ExpectedLoad.md
4. Existing story spec (if any): ProductSpecification/stories/NN-story-name/NN_StoryName.md
5. Archived drafts: ProductSpecification/Archived/DraftStories/1st-iteration/ (scan for relevant files)
6. A named precedent, if this story has one: the interview.md or decisions/*-decision.md of the specific earlier story whose decision this one extends or contradicts — under stories/ or stories/done/, and most often the latter, since a story with a settled decision is usually closed (.claude/rules/workflow.md, "Resolving a story folder"). Never all stories/*/interview.md — that sweep reconstructs the present from a pile of deltas; item 9 answers that question instead. Asking whether a precedent exists at all has its own permitted form — a keyword-scoped grep across stories/** including done/ (.claude/rules/workflow.md, "Asking whether a non-goal or a rationale exists is not the banned sweep")
7. Existing domain code: Scan backend/domain/src/ and backend/usecase/src/
8. Existing adapters: Scan backend/adapters/*/src/
9. What the areas this story touches already do: the acceptance tests of those areas — test class names, scenario descriptions, Statements. This is the current state, not the story folders (.claude/rules/workflow.md, "Where the Current State Lives"). Only enabled tests count as shipped — one still carrying its disable/skip marker is a pending increment, not shipped behavior. Where the suite is silent about an area, read that area's production code and say so in the interview — items 7-8 for a backend behavior, the frontend source for a UI-only one, which Phase 1 does not otherwise scan; silence is not evidence of absence. In a repo with no acceptance suite yet, say so explicitly and fall back to the story folders
10. Existing test cases doc (if any): ProductSpecification/stories/NN-story-name/tests/01_API_Tests.md

Items 4, 6 and 10 resolve stories/NN-story-name/ first, then stories/done/NN-story-name/ (.claude/rules/workflow.md, "Resolving a story folder"). "Not found" for items 4 and 10 means both came back empty — a closed story's spec and tests are in the archive, and reading them as absent is what turns a revisit into a from-scratch re-interview.

Phase 2: Story Selection

Parse user input to determine target story: match by name or number via ProductSpecification/stories.md, or list available stories for interactive selection. If interview.md already exists, warn and ask whether to regenerate or skip.

Phase 3: Interview

Load .claude/templates/spec/interview-format.md for round structure and adaptive questions.

Phase 4: Generate File

Compile all answers into interview.md using the format rules from the template.

Phase 5: Review & Confirm

1. Show the full content of the generated file
2. Ask: "Does this look complete? Any corrections or additions?"
3. If corrections, update the file
4. Report the file path and suggest running /continue NN next (it dispatches the story spec step)

Rules

• NEVER skip the interview — the whole point is interactive knowledge extraction
• NEVER fabricate information — if you don't know, ask
• ALWAYS show what you already found before asking questions (reduce user effort)
• Keep rounds short (1-3 questions max per round)
• Skip irrelevant rounds (e.g., no external API questions for a pure UI story)
• Preserve user's exact wording for decisions and constraints
• The generated file should feel handwritten, not machine-generated
• Do NOT duplicate information already in other interview.md files

Templates

• .claude/templates/spec/interview-format.md — interview rounds, adaptive questions, output format