The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the subprocess module in scripts like run_eval.py and run_loop.py to execute the claude CLI and other local utility scripts. This is necessary for its core functionality of automating skill testing.
[EXTERNAL_DOWNLOADS]: The skill interacts with the Anthropic API (a trusted vendor) via the anthropic Python package. Additionally, the evaluation viewer loads the SheetJS library from a well-known CDN (cdn.sheetjs.com) to render spreadsheets.
[PROMPT_INJECTION]: The agents/grader.md subagent processes data generated by other skills during testing. While this presents an indirect prompt injection surface, it is a functional requirement of the evaluation process, and the skill includes human review steps in its workflow to mitigate automated decision risks.

skill-creator