The Agent Skills Directory

[SAFE]: No malicious patterns, prompt injections, or persistence mechanisms were detected. The skill follows best practices by providing safety warnings against creating malicious skills.
[COMMAND_EXECUTION]: The skill includes Python scripts that use subprocess to execute the platform's CLI tool (claude) for running evaluations and optimizing skill descriptions. It also uses system utilities like lsof for local process management of the evaluation viewer server. These actions are performed locally and are essential for the skill's development workflow.
[DATA_EXFILTRATION]: The eval-viewer component initiates a local HTTP server bound strictly to 127.0.0.1 (localhost). This server is used to render a local web interface for qualitative review of test outputs and does not transmit data to external third-party domains.
[EXTERNAL_DOWNLOADS]: The evaluation viewer template references a well-known library (SheetJS) from a CDN to render spreadsheet files locally in the browser. This is a standard implementation for specialized data visualization.

skill-creator