ce-planlisted

# Plan **Invocation input:** $ARGUMENTS Produce the smallest durable plan that makes implementation safe. Inspect the repository before asking questions, preserve human ownership of scope and architecture, and do not turn deterministic checks into human attestations. This skill is staged. Load only the file for the stage being executed. ## Runtime Inputs - **Outcome (required):** the requested product or engineering result. Ask one short question only when no usable outcome was supplied. - **Brief (optional):** `brief=docs/briefs/<slug>.md`. A brief is discovery context, not a prerequisite. Revalidate its `.json` sidecar before using it to skip answered intent questions. - **Revision (optional):** `revise:<slug>` or a request that clearly targets an existing registered plan. - **Repository evidence:** project rules, relevant code and tests, existing plans/ADRs, build and verification commands, and direct integration surfaces. ## Execution Contract 1. **Inspect before asking.** Resolve the repository shape, existing decisions, constraints, and affected surfaces before asking the human for information the repository already contains. 2. **Ask only consequential questions.** Ask when an answer changes outcome, scope, feature boundaries, architecture, security acceptance, priority, or a hard dependency. Use at most four questions per call and at most two rounds unless the human explicitly continues discovery. 3. **Keep the Scope Lock.** Planning ma