The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains multiple Python scripts that interact with the file system.
scripts/init_skill.py: Creates new directories and writes multiple template files (Markdown and Python) to the local disk based on user-supplied names and paths.
scripts/package_skill.py: Iteratively reads the contents of a directory and creates a compressed ZIP archive (.skill file).
scripts/init_skill.py and scripts/package_skill.py use file system permissions (chmod 0o755) to ensure generated scripts are executable.
[PROMPT_INJECTION]: The skill handles untrusted data that could influence the agent's actions through script parameters and file content generation.
Ingestion points: User-provided command-line arguments (skill-name, --path) in init_skill.py and package_skill.py, and user-authored YAML frontmatter processed by quick_validate.py.
Boundary markers: The skill documentation (SKILL.md) suggests using templates and examples to guide the agent, but the scripts themselves do not enforce boundary markers for path inputs.
Capability inventory: The skill has the capability to read any directory content (via rglob) and write/overwrite files in specified paths.
Sanitization: scripts/quick_validate.py performs regex-based sanitization on the 'name' field (hyphen-case) and 'description' field (length checks and exclusion of angle brackets) to prevent some injection vectors, but it does not sanitize the underlying system paths provided to the script logic.

skill-creator