Self-improve

One light lap: discover → pick (with operator) → implement → review → handoff → STOP. All edits happen in the main checkout.

No autonomous PR merging, no overnight ship loops, no merge cleanup via clock-out, no watch-pr. Git stays operator-owned: .agents/rules/operator-owned-git.md.

Work sources: .agents/context/work-sources.md (incl. read-only .shopping_list.json).

Lap

flowchart TD
  start[Session /self-improve]
  start --> findWork["find-work (read-only)"]
  findWork --> pick[Operator picks Launch N]
  pick --> impl[implement-change]
  impl --> review[review-loop]
  review --> ship["ship-work (handoff only)"]
  ship --> stop[STOP]

- [ ] 1. Run [`find-work`](../find-work/SKILL.md) (read-only)
- [ ] 2. Operator picks a launch brief (or names a task)
- [ ] 3. [`implement-change`](../implement-change/SKILL.md)
- [ ] 4. [`review-loop`](../review-loop/SKILL.md)
- [ ] 5. [`ship-work`](../ship-work/SKILL.md) handoff
- [ ] 6. STOP — do not loop, merge, or clock-out-merge-cleanup

Fuzzy briefs → alignment, then stop for proceed. Protected paths → confirm before editing.

Picking

Situation	Action
Operator names Launch N / a task	Do that
Operator only ran find-work	Stop after briefs
Vague shopping-list title	Alignment / scope first
Dedupe blocked or protected-path	Stop; surface why

Do not walk briefs autonomously into commit/PR. Do not start the next find-work lap until the operator asks.

Do not

• Commit, push, open PRs, or create branches
• Assume overnight Cursor Automations or merge cleanup
• Write .shopping_list.json or .storage/
• Treat ship-work as commit/push — it is handoff only
• Run clock-out as merge teardown (session wrap only, if asked)