Autoreview defines what an independent code reviewer should examine and how to interpret its result. Its repository documentation distinguishes local changes, a committed branch and a single commit, with an explicit severity threshold for findings.
That scope matters when a review reports success. A clean result applies only to the selected changes and severity level; an incomplete assessment or unavailable reviewer cannot count as clean. The documented workflow also calls for checking findings against the code before acting. Its useful contribution is a bounded, interpretable review process.
Terms in this piece · Glossary
agent skill — A reusable instruction file that teaches an agent how to do one job well — the procedure, the tools, and what counts as done.
