The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's core scripts (run_eval.py, improve_description.py, run_loop.py) utilize subprocess to call the claude CLI tool. This is the primary mechanism for executing test queries and performing description optimization within the platform's environment.
[COMMAND_EXECUTION]: The generate_review.py utility uses the lsof command to identify and terminate existing processes on its target port (default 3117) before starting the local evaluation viewer.
[EXTERNAL_DOWNLOADS]: The viewer.html file includes a script tag pointing to cdn.sheetjs.com to load the SheetJS library. This is a well-known technology used locally within the viewer to render spreadsheet outputs (.xlsx files) for the developer to review.
[DATA_EXPOSURE]: The evaluation viewer starts a local HTTP server on 127.0.0.1 to allow developers to review outputs. It reads files from the specific iteration workspace directory and embeds them (including images and text) into the served HTML. This data handling is scoped to the local development environment.

skill-creator