The Agent Skills Directory

[SAFE]: The skill provides instructions for generating well-formatted markdown content, including requirements for frontmatter, structure, and diagrams. It does not perform any dangerous operations or request excessive permissions.
[DATA_EXFILTRATION]: The skill includes a cross-referencing workflow that searches for and reads markdown files in the workspace. This behavior is scoped to project documentation and is necessary for maintaining metadata-based relative links, posing no risk of sensitive data theft.
[COMMAND_EXECUTION]: The skill explicitly instructs the agent not to manually execute linting or package commands, instead relying on built-in platform hooks for validation, which reduces the surface area for command injection.

md-writer