The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes validation scripts including run-tests.sh, quick_validate.py, and validate-agent-plugin.py to check skill structure and specification compliance.
[EXTERNAL_DOWNLOADS]: Utilizes uv run to dynamically load standard packages (jsonschema and PyYAML) at specific versions from the official registry to perform manifest validation.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect injection as its primary function is to ingest and process data from other skills in the repository.
Ingestion points: Reads SKILL.md frontmatter, instruction files (AGENTS.md, CLAUDE.md), and script contents from the skills/ directory.
Boundary markers: None explicitly defined in the parsing logic; uses standard shell utilities and Python scripts to evaluate structure.
Capability inventory: Includes file system read access, shell command execution for syntax checking (bash -n), and metadata extraction.
Sanitization: Employs static analysis tools and exit-code validation to verify the integrity of external skill content.

skill-tester