A developer released AI Slop Detector v2.6.1, a tool designed to identify convincing but empty code generated by AI. The author conducted a self-audit to test the tool's credibility, moving dependency rules into YAML and hardening tests to 165 cases with 85% coverage. This 'trust release' aims to make the detector harder to regress and easier to trust in real-world workflows.

The core experiment involved planting three intentionally flawed files inside the repository to see if the detector would flag them correctly. An empty 0-line file scored a perfect 100, while a file with structural issues like catch-all errors scored 71.21. A third file loaded with buzzwords like 'transformer' and 'optimized' scored 96.77, proving the tool catches both emptiness and misleading language.

Most static analysis tools check for syntax, safety, and complexity. This tool addresses a newer problem in AI-driven development: determining if meaningful implementation exists at all. By offering CI Gate modes—soft, hard, and quarantine—it provides teams with a flexible review signal. The goal isn't to replace human judgment, but to turn vague suspicions about AI-generated code into measurable, actionable follow-ups for reviewers.