python-common-practice
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for an "Inner verification loop" that requires the agent to execute commands (
uv sync,uv run pytest,uv run ruff, etc.) based on the target repository's configuration. This creates an attack surface where a malicious repository could craft apyproject.tomlfile or test suite to achieve code execution when the agent attempts to verify the code. - Ingestion points: The agent is instructed in
SKILL.mdto readpyproject.tomland current repository guidance to determine which tools to run. - Boundary markers: There are no instructions to validate the safety of the repository's configuration or to run tools in a restricted environment.
- Capability inventory: The skill explicitly expects the agent to use shell execution capabilities to run package managers and testing frameworks.
- Sanitization: No sanitization or verification of the repository-provided commands or configuration is performed before execution.
Audit Metadata