The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The evaluation viewer component (eval-viewer/viewer.html) loads the SheetJS library from a well-known CDN (cdn.sheetjs.com) to enable spreadsheet rendering for evaluation results.
[COMMAND_EXECUTION]: The skill uses Python's subprocess module to execute system commands. Specifically, scripts/run_eval.py invokes the claude CLI to test skill triggering behavior, and eval-viewer/generate_review.py uses lsof to manage local server ports.
[REMOTE_CODE_EXECUTION]: The description optimization module (scripts/improve_description.py) interacts with the Anthropic API to generate improved skill descriptions based on test results.
[DATA_EXFILTRATION]: While the skill reads local project files and sends skill descriptions to the Anthropic API for optimization, this behavior is documented and strictly aligned with its primary function as a skill-creator tool.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it processes user-supplied test queries from evals.json and passes them to subagents for grading and analysis. However, this is a necessary part of its function as an evaluation tool.

skill-creator-pro