AI-written PRs can look fine until the risky bits hide in the diff.

I made Agent Diff Sentinel, a small CLI that scans a git diff before merge and flags things worth checking first: auth/billing/security changes, secret-looking lines, missing tests, dependency drift, oversized diffs, and temporary markers.

Repo: https://github.com/subhanA-UA/agent-diff-sentinel

Live page: https://subhana-ua.github.io/agent-diff-sentinel/

Curious how other builders review AI-assisted PRs right now.