AUTO-UPDATED

ProofRun – a local verification receipt for AI coding agents

ProofRun is a new open-source tool that cryptographically binds test results to specific code states, ensuring developers can verify if their software tests remain valid after changes.

Key Points

  • ProofRun uses HMAC-SHA256 signatures to link test outcomes to a specific git commit and file fingerprint.
  • The tool categorizes test statuses as PASS, FAIL, STALE, or NOT RUN, automatically flagging results as STALE if code changes occur.
  • It operates entirely offline without telemetry, relying on real subprocess execution rather than AI-based inference or string matching.
  • The project includes a GitHub Action that independently re-runs tests to prevent reliance on potentially tampered local receipt files.
  • ProofRun is designed to hold AI coding agents accountable by ensuring their claims of passing tests are backed by verifiable, recent execution data.

Why it Matters

This tool addresses the critical "trust gap" in AI-assisted development where agents may falsely claim code passes tests without actually running them. By providing a tamper-evident audit trail, ProofRun helps engineering teams maintain code integrity and prevents the silent introduction of bugs during automated development workflows.
Github.com Published by yebiguo
Read original