The Agent Skills Directory

[COMMAND_EXECUTION]: The orchestration script generate_report.py executes other internal Python scripts using the subprocess module. Evidence: subprocess.run([sys.executable, str(script_path), project_path], ...) in scripts/generate_report.py. Note: The implementation uses a list-based argument structure and avoids shell=True, which mitigates potential command injection from the user-supplied project path.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes content from arbitrary project directories provided as input, creating an attack surface for indirect prompt injection. Ingestion points: The entire directory tree and supported code files are scanned to extract names, structures, and dependencies. Boundary markers: Output is formatted as JSON or Markdown, but there are no specific delimiters or instructions informing a downstream agent to ignore instructions embedded within the extracted data. Capability inventory: The skill possesses the ability to read arbitrary files within the target path and execute local scripts. Sanitization: While results are serialized into JSON, project-derived strings (such as module names and file paths) are included in the generated reports without content-level filtering or sanitization.

project-reader