The Agent Skills Directory

[SAFE]: The skill provides a logical framework for distilling content into reusable instructions. It includes comprehensive guidance on 'suitability tests' to prevent over-automation of high-risk tasks (legal, medical, financial).
[SAFE]: Frontmatter configuration is used correctly to define the skill's name and description without requesting additional tool permissions or network access.
[SAFE]: The reference files (capability-ladder.md and skill-authoring-checklist.md) provide educational content on skill architecture and do not contain executable scripts or remote downloads.
[INDIRECT_PROMPT_INJECTION]: The skill inherently processes untrusted user-provided content to generate skills. While this presents an ingestion surface, the risk is assessed as safe because the skill provides explicit instructions to agents on how to 'strip one-time facts' and 'discard one-time noise,' which acts as a form of manual sanitization during the distillation process. It also mandates the creation of 'should-not-trigger' test cases to harden the resulting skills against unintended behavior.

content-to-skill