The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill contains Python scripts (scripts/validate-component.py and scripts/validate-skill.py) meant for static analysis of skill files. These scripts utilize standard Python libraries such as re, json, and pathlib to lint file content. They do not invoke shell commands with untrusted input or perform unsafe dynamic evaluation of file contents.
[EXTERNAL_DOWNLOADS]: The documentation in references/skill-anatomy.md and references/validation.md refers to using standard package management tools like pip and uv to install dependencies, which is common and expected practice for development toolkits.
[INDIRECT_PROMPT_INJECTION]: The skill's validation tools represent a potential surface for indirect prompt injection as they ingest external markdown and JSON files. However, the ingestion points are limited to local file reads, and the scripts utilize strict parsing (regex and JSON) for structural checks rather than interpreting the file content as instructions. Capabilities are restricted to file reading and console output, following the principle of least privilege for static analysis tools.

meta-skill-creator