m_planlisted

# m_plan Plan → execute → verify. Right-sizes the plan to the task, runs blocker questions in a single batch, executes by default (no `/goal` ceremony unless asked). **Done is decided by `09_verification.md`, not by the agent's judgement.** ## Inputs - `$ARGUMENTS`: task description. If empty, ask once and wait. ## Phase 0 — Blockers + sizing (one round) 1. **Scan in parallel** (read-only): `git status`, `git log --oneline -10`, `CLAUDE.md`, `AGENTS.md`, `.claude/rules/*.md`, `infra.md`, `deploy.md`, top-level tree, package/build manifests, existing `.m_plan/`. For a large or unfamiliar repo, delegate the codebase portion of this scan to the `m_code-context-scout` agent and keep only its map. 2. **Pre-classify task size** by these classes: - **tiny** (one file, < 50 LOC, no infra, no migration) → files `01, 04, 05, 09`. - **small** (one module, tests required, no infra/deploy delta) → files `01, 04, 05, 07, 09`. - **medium** (multi-module, deploy delta, no infra change) → files `01, 02, 04, 05, 06, 07, 09`. - **large** (touches infra, migrations, or external contracts) → all 9. - **`08` is added on top of ANY size** when the task adds an externally-callable or browser-observable surface (HTTP endpoint, UI page, CLI command, Telegram command) that integration tests can't fully cover — this is the one artifact that crosses size bins (see the Phase 1 table). So a *small* UI change is `01, 04, 05, 07, 08, 09`. Skipped artifacts are **not written** — `09_ver