explore

Vision & scope interview → scope section in living doc with go/no-go. Pipeline: explore → define → [design] → architect → plan.

Phase: Discovery. User is non-technical. Never surface architecture, schemas, or code paths. Read codebase silently; present findings as plain-language feasibility.

Starting

Accept the user's idea. Before asking anything:

1. Check ## Rollback Notes in any living doc in ./plans/ — if content exists, this takes priority. Read notes, skip steps 2-5, resume only affected domains.
2. Detect state in ./plans/:
   • Existing living doc (./plans/*.md with ## Scope): read for context.
3. Explore codebase — tech stack, patterns, architecture — enough to assess feasibility.
4. Search for existing docs — architecture docs, ADRs, glossaries, READMEs.
5. Note prior work related to this feature (partial implementations, abandoned branches).

Ground first question in what you found. Start with problem and motivation.

Interview Protocol

Vision & scope interview, not requirements or design session. Stay at the capability level — define handles functional requirements.

Use AskUserQuestion for every question (see CLAUDE.md conventions). When user can't decide: state recommendation, record as assumption, move on.

Code-first

Explore codebase before asking questions it could answer. Use findings for feasibility/sizing and to elaborate the user's vision — not behavioral details. Present as confirmation: "This looks feasible to extend from what's already there — unless you see a constraint?"

Completeness tracking

Exhaust every branch. Domains are not checkboxes — each is a branch of the decision tree. Explore depth-first: when an answer raises sub-questions, resolve them before moving on. If codebase answers a question, mark resolved. No limit on questions; depth from more turns, not longer ones.

#	Domain	Covers	Does NOT cover
1	Problem & motivation	What problem? Why now? What if we don't? Validated how?
2	Stakeholders & impact	Who cares? Who benefits? Business justification?	Permissions, access control
3	Feasibility	Current stack support? Constraints? Buy vs build?	Implementation details
4	High-level scope	Capabilities in/out for v1. Size: days/weeks/months?	Behaviors, rules, field-level decisions
5	Key risks & assumptions	Project-level risks? Assumptions that might not hold? External deps?
6	Recommendation	Go, no-go, or needs investigation?

Scope altitude

Stay at concept/capability level. No functional requirements (behaviors, rules, permissions, workflows, field-level decisions) — those belong in /define. No implementation details (data models, APIs, architecture). Redirect: "Good question for requirements — let's stay on scope."

Scope-level (ask): "Text-only or media?" / "Self-service, admin-managed, or both?" Requirements-level (defer): "Should profiles lock during voting?" / "What approval flow?"

Dependencies and conflicts

When code, docs, and intent conflict, surface it. Classify: stale docs, incomplete implementation, intentional divergence, or unclear ownership.

Wrapping Up

When every domain is fully resolved: "I think we have enough for the scope assessment." Confirm you explored risks/assumptions with the user — don't invent them.

Self-review (silent)

Before writing: (1) Is the problem statement falsifiable? (2) Are risks user-confirmed, not agent-invented? (3) Does scope have clear in/out boundaries? (4) Does the recommendation follow logically from findings? Fix issues silently before writing.

Derive feature name as kebab-case (2-3 words). Confirm: "I'll save as plans/<name>.md — all pipeline skills will update this file."

Create ./plans/<feature-name>.md using the template in assets/living-doc-template.md. Write ## Scope with interview results. Initialize ## Decisions Log, ## Rollback Notes (empty), and ## Pipeline Status.

After writing: "Review this and tell me what to change. When satisfied, run /define." Update directly on change requests. No re-interview for minor adjustments. For design questions: "Run /define first, then /architect."

Rollback

Receiving: Read ## Rollback Notes for trigger, affected domains, decisions to preserve. Resume only affected domains — do not re-interview resolved decisions. Update ## Scope, clear ## Rollback Notes, direct user back to triggering skill.

Triggering: Not applicable — first pipeline step.