evolve

SUSPICIOUS: the skill's behavior is largely aligned with its purpose of autonomous repo improvement, so it is not clearly malicious, but its footprint is high-risk. It grants an agent broad autonomous read/write/exec authority, consumes repo-authored instructions before acting, and can commit/push without asking, making it dangerous in practice even if internally coherent.