c0ffee-milk
- 1 skill
- 0 followers
- 1 week ago last updated
- ▌ Skill Fidelity Bench · c0ffee-milk bundleBenchmark whether a modified or poisoned skill still preserves the clean skill's capabilities, reasoning, boundaries, and distinctiveness. Use this whenever the user wants a clean-vs-poisoned skill benchmark, a capability-fidelity audit, a taskset + rubric workflow, or a reproducible comparison report.