The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Python's subprocess module to orchestrate the evaluation workflow. It executes the claude CLI for automated testing and optimization, and uses system utilities like lsof and kill to manage the local web server's lifecycle. These actions are transparently documented as part of the tool's core functionality.
[EXTERNAL_DOWNLOADS]: The evaluation viewer (viewer.html) includes a reference to the SheetJS library hosted on cdn.sheetjs.com. This well-known service is used to render spreadsheet data within the local review interface.
[DYNAMIC_EXECUTION]: To test new or modified skills, the run_eval.py script dynamically generates temporary command files within the .claude/commands/ directory. This mechanism allows the agent to interact with the skill-under-test in a realistic environment.
[DATA_EXFILTRATION]: The skill launches a local HTTP server on localhost to host the evaluation viewer. This server is used exclusively to present gathered metrics and generated outputs to the user for qualitative assessment and does not transmit data to external servers.

skill-creator