The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/analyze_all.py coordinates the analysis by invoking specialized sub-scripts using subprocess.run. The execution is handled safely by using sys.executable and passing arguments as a list, which prevents shell injection vulnerabilities.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it processes untrusted Python source code provided by users. Ingestion points: Multiple analyzers in the scripts/ directory read untrusted code files via Path.read_text(). Boundary markers: The tool outputs structured JSON reports to organize its findings, but it does not include explicit instructions or markers to prevent the agent from following commands embedded in the code comments. Capability inventory: The orchestration script uses subprocess.run for internal task management. Sanitization: The input code is parsed into an AST for structural analysis, but no filtering or sanitization of string literals or comments is performed.
[SAFE]: The tool relies on static analysis via the ast module (e.g., in analyze_complexity.py). This design ensures that the code under review is never actually executed, effectively neutralizing runtime threats from malicious source files.
[SAFE]: No indicators of data exfiltration, hardcoded credentials, or unauthorized network activity were found during the analysis of the scripts and documentation.

python-simplifier