The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The skill instructs the agent to execute commands such as uv run pytest and prek run -a. Since pytest automatically executes Python code found in test files, an attacker can achieve arbitrary code execution by placing malicious code in the repository being analyzed.
[EXTERNAL_DOWNLOADS] (MEDIUM): The skill recommends installing prek via uv tool install prek. This tool is not from a verified or trusted source, posing a supply-chain risk.
[PROMPT_INJECTION] (HIGH): This skill is highly susceptible to Indirect Prompt Injection (Category 8).
Ingestion points: Reads and processes the entire Python codebase, including source files and tests.
Boundary markers: None are specified; the tools interact directly with raw file content.
Capability inventory: Possesses the ability to execute subprocesses (ruff, pytest, prek) and modify local files (ruff --fix).
Sanitization: No sanitization or safety checks are performed on the repository content before execution.
[REMOTE_CODE_EXECUTION] (HIGH): The combined risk of running a tester (pytest) on untrusted code and executing an untrusted third-party binary (prek) facilitates remote code execution on the environment where the agent operates.

python-quality-tooling