The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Python subprocess module to execute system commands and CLI tools. Specifically:
scripts/improve_description.py and scripts/run_eval.py call the claude CLI to process prompts and test triggering logic.
eval-viewer/generate_review.py uses lsof to manage port availability for its local web server.
These operations are essential for the skill's primary function as a development and benchmarking utility.
[REMOTE_CODE_EXECUTION]: The skill's core loop involves generating prompts based on performance data and sending them to an external LLM (via the claude CLI) to improve code or descriptions. While this involves remote logic processing, it uses the platform's official tools and is the intended primary purpose of the skill.
[DATA_EXPOSURE]: The eval-viewer component starts a local HTTP server (HTTPServer) on localhost to provide a GUI for reviewing evaluation results. This server is scoped to the workspace directory and provides a mechanism for the user to review local files and provide feedback, which is standard for developer-centric tools.

skill-creator