deep-planlisted

<!-- prettier-ignore-start --> **Document Version:** 3.0 **Last Updated:** 2026-03-07 **Status:** ACTIVE <!-- prettier-ignore-end --> # Deep Plan Structured discovery-first planning that produces high-quality implementation plans by front-loading decision-making through exhaustive questioning. ## Critical Rules (MUST follow) 1. **No artificial caps** — no ceiling on questions or plan size. Floor of ~15 questions. Plans are as long as they need to be. 2. **Persist state incrementally** — save decisions to state file after every batch. Long planning sessions WILL hit compaction. 3. **Every plan includes audit checkpoints** — at minimum at plan completion, at phase boundaries for multi-phase plans. 4. **Show plan inline** — users may not have filesystem access. Write to file for persistence, display in conversation as primary output. 5. **DECISIONS.md is standalone** — not embedded in the plan. Most-referenced artifact during execution. 6. **Handoff is a decision tree, not a phase** — after approval, route to the right execution approach (subagent / GSD / manual). 7. **PLAN.md hygiene is a live contract** — plans MUST be updated as work completes: top status banner (`X / N steps ✅`), per-step header marker (⏳/🔄/✅/❌/⏸), and Status Summary table MUST reflect reality. Same hygiene contract as TODOS.md ↔ todos.jsonl. `npm run plans:hygiene` flags drift; session-end surfaces it. Drift discovered post-hoc (as with T28 + T29 PLANs in Session #279)