The Agent Skills Directory

[SAFE]: The skill functions as a purely informative reference for the agent to understand the structure and conventions of skill files. It is configured as user-invocable: false and internal: true, limiting its scope to internal guidance.
[SAFE]: Documentation regarding security-sensitive features, such as dynamic context injection (!command) and tool permissions (allowed-tools), is presented descriptively for educational purposes. The file explicitly avoids using the actual execution syntax to prevent accidental command execution during loading.
[SAFE]: All external links point to official documentation sites (code.claude.com, platform.claude.com) or the official Agent Skills standard domain (agentskills.io), which are trusted sources.

doc-skills