What it does
It reads the local transcript for evidence that verification happened after the final changes, then allows or blocks the stop.
Agent tooling
jev-belay
A Claude Code stop hook that checks whether a claim of finished work is backed by later passing checks.
Bookmark: jev-belay
Keep this in your collection.
Leave a noteWhat would you try with this?
: jev-belay
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
Your developer can add this tool to an AI assistant called Claude. The tool looks at what Claude just typed. It checks if Claude changed files without verifying its work. If so, it sends the final message to an outside AI service called TypeSafe.
Your developer needs an access key from TypeSafe so the tool can reach it. The service decides if Claude wrongly claims the job is done. If it does, the tool forces Claude to go back and check the files. If the connection breaks, the tool lets Claude finish normally.