The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides explicit guidance on safe process execution.
Recommends subprocess.run with list-based arguments over shell=True to prevent command injection in references/anti-patterns.md.
Warns against using os.system and string-concatenated paths.
[INDIRECT_PROMPT_INJECTION]: The skill addresses potential injection risks through structured output and validation.
Provides predefined templates in assets/templates/ to ensure a consistent security posture.
Includes scripts/validate-script.py to programmatically verify that generated scripts follow safety rules like specific exception handling and proper shebangs.
[EXTERNAL_DOWNLOADS]: Manages external dependencies using established ecosystem tools.
Supports standard package managers like pip, uv, and poetry as documented in references/package-managers.md.
Suggests well-known, trusted libraries such as tqdm for progress bars and PyYAML for configuration handling.

python-script-developer