codebase-auditlisted

# Audit a codebase into a ranked what-to-do list The advise-only front-end for an existing codebase. Point it at a repo or an area and it returns one thing: a leverage-ranked list of what is worth doing, each finding grounded in repo evidence, with the handoff being a human who picks one item and enters `/plan-grill` or `/diagnose`. One job: **surface what is worth doing, ranked and evidenced, and hand off - never build.** This is the entry the other front-ends don't cover: `/plan-grill` needs a chosen feature, `/diagnose` needs a symptom, `/groundwork` is for greenfield. This one starts from "here is a codebase, tell me where to spend effort." When a feature or a symptom is already named, route there instead - auditing a repo to answer a question you can already state concretely is wasted motion. Read `.better-dev/overrides.md` first - a project override (an area to always skip, a house severity language, a "we never audit vendored code" rule) wins over anything here. ## 1. Orient before you judge Get structure first: compose `/codebase-map` for callers, dependents, the real schema, and the blast radius of the areas you will look at. An audit that guesses at structure surfaces findings at the wrong seam. Ask the leverage question structurally before reading anything: `/graphify-wrapper-query <name> --hubs` lists the most-connected nodes, which is this skill's own "where is the leverage" rendered from the graph rather than from a sweep. A hub with weak tests, no error h