refine-ideaslisted

# /refine-ideas $ARGUMENTS Optimize IDEAS.md by surfacing the ONE irreducible question: "Why is this necessary?" Groups items by answer, flags unclear ones, shows prioritization without enforcing format. **Synergy:** `/prioritise` for pure ranking of an already-clear list (no grouping narration, just ranked output with justifications). Use `/refine-ideas` first when priorities are unclear; hand the output to `/prioritise` to rank within or across groups. `$ARGUMENTS` = repo name or path. If none provided, uses current directory. ## Core Principle **Good IDEAS.md = clarity on "why each item matters."** Tiers, formats, validation checklists are scaffolding. The real work is answering: "If we skip this item, what fails?" **This skill does NOT impose a format.** It works on any existing structure (Pending/Future, Phase 1/2, TIER 0-4, unstructured narrative). It derives priorities FROM the why-necessary answer, not from tier labels. --- ## Phase 1: Locate repo & read IDEAS.md 1. Resolve repo path (check filesystem or use current directory) 2. Read `<repo>/docs/IDEAS.md` 3. Extract all items/ideas (any format) --- ## Phase 2: For each item, extract or ask ONE question **Question:** "Why is this necessary? What fails if we skip it?" For each item: - If answer is already clear from the text -> extract it - If answer is vague or missing -> ask Claude to infer it from context, flag as "NEEDS CLARIFICATION" --- ## Phase 3: Group items by answer type Create 5 groups (not