The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions and templates utilize the uv run command to execute generated Python scripts.
[EXTERNAL_DOWNLOADS]: The PEP 723 metadata in the template allows for automatic installation of Python packages from official registries like PyPI when the script is first run.
[INDIRECT_PROMPT_INJECTION]: The template references/uv-script.template.py includes placeholders that interpolate user or agent content into an executable script, creating a potential surface for instructions to be interpreted as code.
[INDIRECT_PROMPT_INJECTION] Mandatory Evidence Chain: (1) Ingestion points: Placeholders like {{MAIN_LOGIC}}, {{SCRIPT_DESCRIPTION}}, and {{DEPENDENCY_1/2}} in references/uv-script.template.py. (2) Boundary markers: None identified in the template. (3) Capability inventory: Supports script execution and file modification via the uv CLI. (4) Sanitization: No input validation or escaping is applied to the content injected into script placeholders.

python-scripting