The Agent Skills Directory

[COMMAND_EXECUTION]: The script_tester.py script uses subprocess.run to execute Python scripts found within the directory path provided by the user. While this is the intended purpose for testing, it creates a vector for arbitrary code execution if the tool is pointed at a malicious directory.
[EXTERNAL_DOWNLOADS]: The scripts quality_scorer.py and skill_validator.py import the yaml module (PyYAML). This is an external dependency that is not part of the Python standard library, which contradicts the documentation's claim that the skill has "Zero External Dependencies".
[REMOTE_CODE_EXECUTION]: The testing framework facilitates the execution of local files through the script_tester.py utility. In an autonomous agent context, an attacker could potentially trick the agent into running malicious code by directing the tester to a compromised path.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and processes untrusted files from the file system without sufficient isolation.
Ingestion points: The skill_path argument defines the source for scripts, metadata, and documentation files processed by the tool.
Boundary markers: None. The tool reads and analyzes file content directly without markers or instructions to ignore embedded logic.
Capability inventory: subprocess.run in script_tester.py provides code execution capabilities. Path.read_text across all scripts allows file system access.
Sanitization: Minimal; the scripts use Path.resolve() for path validation but do not sanitize the content of the files before analysis or execution.

skill-tester