skill-creator

To create a new AI skill from scratch, you define the skill's purpose and expected output, then the system generates a SKILL.md file using Python and Markdown definitions. This interactive process captures intent and structures the skill configuration.

You can improve an existing AI skill by processing user feedback and benchmarking results. The system iteratively modifies skill definitions, runs test cases, and compares performance against baselines to optimize behavior.

The best way to measure AI skill performance is by executing test cases and comparing results against baselines. The system uses subagents for parallel evaluation, providing detailed metrics on skill accuracy and behavior.

Yes, you need a Python environment with specified libraries and tools to run skill creation and evaluation workflows. The system depends on subagents, eval-viewer, and glm components to facilitate parallel execution and interactive design.

Yes, you can use subagents for parallel execution when testing AI skills. The system utilizes subagents to run concurrent evaluations, enabling efficient measurement of skill performance across multiple test cases.

Markdown-based definitions for skill improvement rely heavily on a specific Python environment and dependencies like subagents and glm. Limitations include potential bottlenecks in interactive design if the required libraries are not properly configured.