Skills
A constraint language over the interface caught 30 of 30 injected bugs with zero false positives where three frontier models could not
FlowCheck lets someone specify user-visible information flows through the application interface rather than the code, then compiles those constraints into deterministic CodeQL analyses. Across four applications generated with Claude Code it correctly translated and flagged all 30 injected constraint violations with no false positives, while Claude Opus 4.7, DeepSeek V3, and Gemini Pro prompted to find bugs in the same code all fell well short of full accuracy. The pattern worth stealing is stating intent in interface terms and checking it deterministically, rather than asking a model to review code it just wrote.
↳ Follow the thread