Verifies skill reliability and pressure resistance through a quantitative TDD cycle and compliance scoring.
Ensures the robustness of Claude Code skills by subjecting them to a rigorous Red-Green-Refactor validation process before deployment. This framework simulates high-pressure environments—such as tight deadlines, authority constraints, and sunk costs—to identify and close loopholes in agent behavior. By generating quantitative Serena compliance scores and tracking loophole counts, it transforms qualitative process documentation into measurable, bulletproof implementation patterns that resist rationalization.
主要功能
01Pressure scenario simulation for time, authority, and sunk cost constraints
02Quantitative compliance scoring (0.00-1.00) based on scenario performance
03Serena metrics integration for historical pattern learning and threshold tracking
041 GitHub stars
05Red-Green-Refactor cycle adapted for agent behavior validation
06Loophole identification and explicit negation to prevent agent rationalization
使用场景
01Auditing how agents handle conflicting priorities like speed versus quality
02Optimizing skill clarity by identifying and removing vague instructions
03Hardening discipline-enforcing skills like TDD or code review before deployment
