Open source tool for coding agents

show-receipts

Your agent says the tests pass. This makes it prove it.

A Stop hook for Claude Code and Codex. When the agent is about to end its turn with "all tests pass", "the build succeeds" or "lint is clean", the hook reads the session transcript and looks for a matching command that ran after the last code edit and exited 0. If there is none, the stop is blocked and the agent is told which sentence has no receipt. No model judges the model: it is one file of standard library Python that never calls the network, and it leans hard toward letting things through, because a wrong block is worse than a missed claim.