ProofRun Verifies AI Coding Agent Test Outcomes
TL;DR. ProofRun introduces a verification tool that cryptographically confirms if AI coding agents truly ran tests and passed them against the exact code state. - The tool binds test results to a precise code fingerprint, including git commit and uncommitted changes. - If code changes, previously recorded test results automatically become 'STALE', preventing false positives from agents. - ProofRun operates offline, uses no AI for verification, and directly executes subprocesses for concrete results.
- ProofRun verifies test results from AI coding agents.
- It uses cryptographic fingerprinting to link test outcomes to specific code states.
- The tool identifies 'STALE' results if code changes after a test run.
- ProofRun works offline and does not use AI for its verification process.