The Agent Skills Directory

[DYNAMIC_EXECUTION]: The scripts/benchmark_changes.py script utilizes importlib to load and execute Python modules from arbitrary file paths provided via command-line arguments. While this is the intended mechanism for performance benchmarking of refactored code, it constitutes dynamic execution of potentially untrusted local code.
[COMMAND_EXECUTION]: Multiple scripts, including scripts/analyze_multi_metrics.py and scripts/analyze_with_flake8.py, use the subprocess module to execute external CLI tools like flake8, radon, and complexipy. These tools are used for automated code quality and complexity analysis.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and refactor untrusted Python source code, creating an indirect injection surface. 1. Ingestion points: Local Python files are read and parsed by AST and metric calculation scripts. 2. Boundary markers: No explicit delimiters or safety instructions for the ingested code are present in the skill instructions. 3. Capability inventory: The skill has the capability to write to the filesystem, execute subprocesses (linters/tests), and dynamically load modules. 4. Sanitization: No sanitization of the input source code is performed prior to analysis or execution.

python-refactor-method