The Agent Skills Directory

[SAFE]: The skill provides instructions for creating Agent Skills using templates and checklists. No security vulnerabilities were identified in its design or instructions.
[COMMAND_EXECUTION]: The skill suggests the execution of a local tool skills-ref to validate the created skill's directory and frontmatter. This is a standard part of the skill development lifecycle.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection due to processing untrusted external source materials to generate skill instructions. 1. Ingestion points: Domain knowledge, task traces, and project documentation (SKILL.md). 2. Boundary markers: No explicit markers used to separate untrusted source material. 3. Capability inventory: File creation and local command execution (skills-ref). 4. Sanitization: No input validation is performed on the source material.

skills-creator