venv-manager

Pass

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the subprocess module to execute environment management tasks, including pip install, ensurepip, and running user-specified scripts from other skills. These operations are restricted to the context of the managed virtual environments.
  • [EXTERNAL_DOWNLOADS]: In cases where ensurepip is unavailable, the skill is configured to download the get-pip.py utility from https://bootstrap.pypa.io/. This is an official, well-known service for Python packaging tools.
  • [REMOTE_CODE_EXECUTION]: The skill executes the get-pip.py script downloaded from the trusted PyPA domain to install the pip package manager within a newly created virtual environment.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because parser.py extracts dependency lists and group names from the SKILL.md files of other potentially untrusted skills. This information is used to define installation targets and file system paths for virtual environments.
  • Ingestion points: Reads SKILL.md and requirements.txt from other skill directories.
  • Boundary markers: None implemented for the parsed dependency data.
  • Capability inventory: Uses subprocess.run to install packages and venv.create for environment setup.
  • Sanitization: Package names are extracted via regex and split into lists for subprocess.run, which provides some protection against shell injection, but group names are used directly in path construction without validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 15, 2026, 11:51 AM