plan-trimlisted

# plan-trim Investigate, propose cuts, confirm, then execute. Be terse. Don't ask clarifying questions before the proposal — investigate first. ## Phase 1: Investigate Read primary sources, not just the plan index. The cuts only land if you can name the unexamined assumption inside each phase doc. 1. **Plan index.** Read `documentation/plan.md` (or `plan.md` / `README.md`). Note the status table and doc map. 2. **Recent git activity.** `git log --oneline -30`. Note the most recent shipped milestone — its scope often subsumes pending plan items. 3. **Open phase docs.** Read every phase doc not ✅ DONE or ARCHIVED. Hunt for: calendar gates, sub-goals met by another phase, de-facto-exercised UAT, dead cross-refs. 4. **Run-doc archive.** `ls documentation/runs/` (or equivalent). UAT pass docs are evidence ceremonial gates can close. 5. **Cross-reference grep.** For each doc you suspect is dead, `grep -r <doc-name>` — the result tells you whether deletion is safe or needs redirects first. ## Phase 2: Propose cuts Output a single bulleted list. Each item names the **unexamined assumption** and **why it might not hold**, not just the item to cut. A cut without an assumption named is a guess; a user can't validate a guess. - **Calendar gates that are vibes.** "≥N months of X before Y" — is N load-bearing or a round number? Replace with the real prereq (data volume, schema, capability). - **Sub-goals shipped elsewhere.** A phase scoped A+B+C may have had B shipped under another