The Agent Skills Directory

[SAFE]: The skill provides instructions and a robust methodology for developing AI agent skills. It includes detailed guides on capturing user intent, writing SKILL.md files, and implementing progressive disclosure patterns.
[SAFE]: The skill incorporates a structured testing and evaluation loop, utilizing subagents to run parallel test cases with and without the skill to benchmark performance and capture timing data.
[SAFE]: The skill includes instructions for descriptive optimization, using a multi-step process to generate trigger evaluation queries and iteratively refine the skill's description to improve triggering accuracy.
[SAFE]: The skill references several internal tools and scripts for benchmarking, grading, and viewing results, which are part of the intended development environment.
[SAFE]: The skill includes explicit security guidelines under the 'Principle of Lack of Surprise,' instructing the model to avoid creating skills that contain malware, exploit code, or facilitate unauthorized activities.

skill-creator