The Agent Skills Directory

[SAFE]: No malicious patterns or unauthorized behaviors were detected. The skill facilitates skill development by automating common tasks.
[COMMAND_EXECUTION]: The skill uses the subprocess module to execute the claude CLI and several internal Python scripts (e.g., run_loop.py, package_skill.py) to run evaluations and manage the skill development lifecycle.
[DATA_EXFILTRATION]: A local HTTP server is hosted on the loopback interface (127.0.0.1) by the generate_review.py script to allow the user to review evaluation results in a web browser.
[PROMPT_INJECTION]: The skill processes user-provided prompts for testing purposes; while this represents a surface for indirect prompt injection, it is managed within a developer-controlled testing environment.

skill-creator