security-reviewlisted

# Reviewing this work item's diff for security `chain_review` added this task because the plan touches something that decides *who may do what* — authentication, session handling, tokens, secrets, or a permission check. You are not repeating `on.review.local.run`'s general review; you are reading the same diff with one question in mind, and you are allowed to find nothing. ## What you are looking at The diff for this work item is handed to you by path, as a review package. Read the files it touches around the change, not just the changed lines. Almost every real finding here lives in the interaction between new code and an assumption the old code was already making. ## What earns a finding A finding is a specific defect at a specific place. "Consider hardening this" is not a finding. If you cannot name the file and say what an attacker or an unlucky caller actually gets, you do not have one. Look for, in rough order of what actually bites: - **A check that does not run.** A permission or ownership test skipped on one path, behind a cache, after an early return, or in the error branch. The most common real hole is not a wrong check — it is a missing one on the second route to the same resource. - **Trust drawn from the wrong place.** A caller's identity taken from a request body, a header the client controls, or a work item's own record rather than the session. Kraft's own model has a live example worth pattern-matching against: a worker must not approve its