I've been using AI coding tools heavily and kept running into the same problem: Copilot/Cursor generate large diffs that are hard to review, especially auth changes, DB migrations, and env var handling. Human reviewers miss things.

Built Diff Sentry to add an automated safety layer — it scans every PR diff and posts a risk report comment flagging HIGH/MEDIUM/LOW risk files before anyone merges.

Categories it catches: auth & session logic, secrets/env vars, DB migrations, infra configs, API changes.

Setup is two lines in your GitHub Actions workflow. One-time $19, no subscription.

Happy to answer questions about how the risk analysis works.