studlisted

# Stud: Part 1, plan then skeleton (command) You were invoked deliberately by the user with an outcome in hand. Your job is to turn a decided-upon feature into a **reviewable runnable skeleton**, and to stop there. This is a planning ritual, not an implementation run: interview first, stud second, then hand back for review. Do **not** write real logic in this skill. **Arguments:** `$outcome` is required; `$questions` is optional (trailing). If `$questions` is empty, skip the interview (Step 1) and go straight to exploring. ## Overview Deliver **$outcome** in **two stages** so it can be understood before it hits maximum complexity: - **Stage 1, stud (this skill):** every function studded in its *real* file/location with real signatures, but bodies return realistic **fake data**. The whole chain runs end-to-end on dummy data. Names, placement, and contracts get reviewed here. - **Stage 2, fill (separate):** replace the fakes with real logic, pure/leaf functions first. Each `should` bullet becomes a test. **Core principle:** shape before logic. A stud that *runs* (fake data, real wiring) is worth more than a doc, because you can execute the plumbing and see the flow before writing a line of real logic. The mechanical conventions (scaffold banners, seam markers, comment-by-language, fill order) live in `~/.claude/library/references/stud/conventions.md`. Read it before Step 3. A full worked example is in `~/.claude/library/references/stud/worked-example.md`. ## Steps ###