ce-plan-auditlisted

# Plan Audit **Invocation input:** Plan to audit (optional): $ARGUMENTS Independently audit the **written plan artifact** of a `docs/plans/<slug>/` directory — for structural integrity, the soundness of its decisions and decomposition, fit to the codebase it was planned against, and any drift between its files. It reports **evidence-backed findings** and **escalates**; it never edits the plan, re-decomposes the project, or renders a go/no-go on the plan. `/ce-plan` validates only the **in-flight candidate draft** and writes the artifact exactly once (its Execution Contract bars writing before final approval). Every `/ce-plan` gate — sizing, reachability, session-fit — runs on the conversation-held draft, so **nothing re-checks the persisted plan** after a manual edit, drift, or a hand-authored plan. This tool does. It also machine-**proves** the dependency-direction and cycle-freedom invariants that `/ce-plan` itself only *model-reads* (its Honest Limitations: "Dependency direction is not machine-proven"). This is the **plan-layer sibling of `review`** (which audits implemented *code*) and the **static counterpart of `retro`** (which aggregates *run-time* metrics and writes nothing). The discipline chain gains a pre-flight auditor on the plan itself: ``` plan ─► { plan-audit } ─► spec ◄── implement ◄── { verify · review } ``` **Independence is the point.** An auditor reading the plan in a fresh context catches what the planning conversation cannot — the same argument b