The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the subprocess module to perform environment management tasks.
Evidence: scripts/venv_manager.py uses subprocess.run to execute ensurepip and pip install commands for package management.
Evidence: scripts/venv_manager.py uses subprocess.call in the cmd_run function to execute user-specified Python scripts within the managed virtual environments.
Evidence: tests/test_cli.py uses subprocess.run to execute the skill's own CLI during automated testing.
[EXTERNAL_DOWNLOADS]: The skill contains logic to download official tooling from a well-known Python infrastructure provider.
Evidence: In scripts/venv_manager.py, if the standard ensurepip module fails, the skill attempts to download get-pip.py from https://bootstrap.pypa.io/get-pip.py to install pip manually. This is a well-known and official source for Python packaging components.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from other skills which could be manipulated by an attacker.
Ingestion points: The skill parses SKILL.md (YAML frontmatter and prose) and requirements.txt files from other skill directories to identify dependencies (found in scripts/parser.py).
Boundary markers: The skill does not implement explicit boundary markers or isolation when parsing these external skill definitions.
Capability inventory: The skill has the capability to write files (creating venvs) and execute commands (pip install, subprocess.call).
Sanitization: The skill uses regular expressions to extract package names but does not strictly validate these names against a whitelist before passing them to the system shell via pip.
[DYNAMIC_EXECUTION]: The skill uses dynamic module loading for internal components.
Evidence: scripts/detector.py and scripts/venv_manager.py use the __import__ function to load the local parser and detector modules. This is used for code modularity rather than executing untrusted external code.

venv-manager