The Agent Skills Directory

[SAFE]: The skill provides a robust framework for performance optimization, strictly mandating baseline measurements, profiling to identify bottlenecks, and re-benchmarking to verify gains. It includes strong guardrails against 'guessing' and requires reverting any changes that are not measurably faster or that break existing functionality. The use of referenced skills like converge-loop and adversarial-verify supports a controlled and safe execution environment. No suspicious patterns such as prompt injection, data exfiltration, or obfuscation were found.

auto-performance