The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided Python files using static analysis scripts. This introduces a potential surface for indirect prompt injection if the files being audited contain malicious instructions designed to influence the agent's refactoring logic or commit messages.
Ingestion points: Python modules targeted by scripts/count_shapes.py and scripts/find_mutations.py.
Boundary markers: The procedure explicitly defines a strict loop (Baseline -> Checkpoint -> Refactor -> Verify) and classifies the test suite into frozen entry-point tests and internal tests to prevent unintended behavior changes.
Capability inventory: The agent is instructed to read local files, execute provided AST-based analysis scripts, modify source code, and perform git commits.
Sanitization: The included scripts use the Python ast module, which parses the structure of the code but does not execute its content.
[COMMAND_EXECUTION]: The skill includes two Python utility scripts (scripts/count_shapes.py and scripts/find_mutations.py) that the agent is instructed to execute locally on the project being audited.
Evidence: python3 <skill>/scripts/count_shapes.py <module.py>... and python3 <skill>/scripts/find_mutations.py <module.py>... instructions in SKILL.md.
Analysis: These scripts use only the Python standard library (ast, argparse, pathlib) to perform static analysis. They do not perform network operations, file writes, or dynamic code execution.
[EXTERNAL_DOWNLOADS]: The skill README provides instructions for installing the skill from a remote GitHub repository using the platform's standard installation command.
Evidence: npx skills add TheodoreAD/agent-skills --global --skill python-refactor-audit in README.md.
Analysis: This is the standard mechanism for adding skills to the agent environment. The resource originates from the skill author's own repository.

python-refactor-audit