The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes Python scripts via heredocs and utilizes system utilities such as sed and pandoc to perform text manipulation, formatting, and file merging operations on Markdown documents.
[EXTERNAL_DOWNLOADS]: The conversion tools reference https://cdn.simplecss.org/simple.min.css to provide styling for generated HTML files. This is a well-known CSS framework CDN and does not pose a security risk.
[DYNAMIC_EXECUTION]: The YAML validation script uses yaml.safe_load() to parse frontmatter content. This is a secure implementation that prevents the execution of arbitrary code embedded in YAML data.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external Markdown files, which serves as an ingestion point for untrusted data. While the scripts process content as raw text, the resulting output could potentially contain embedded instructions intended for the agent.
Ingestion points: Files passed via CLI arguments to processing scripts (SKILL.md).
Boundary markers: None implemented for processed markdown content.
Capability inventory: File reading (open), local file modification (sed -i), and directory listing (glob.glob).
Sanitization: Uses yaml.safe_load() for metadata parsing.

markdown-toolkit