The Agent Skills Directory

[SAFE]: The skill is a well-documented development tool that follows security best practices, including an explicit 'Principle of Lack of Surprise' that forbids the creation of malicious or misleading content.
[COMMAND_EXECUTION]: The skill supports generating and running local Python scripts for automated grading and benchmarking, which is a key part of its functionality as a development tool.
[PROMPT_INJECTION]: The skill processes user-supplied prompts to generate new skill instructions and evaluation criteria. This is the primary intended use case for the assistant, and the skill's workflow includes human-in-the-loop review phases.

skill-creator