rolling-codes/better-skill-creator ? reverse-engineered prompt

Build me a Python tool and Claude Code skill that helps create better skills with real checks instead of just hoping they work.

I want it to guide me through the whole process, from understanding the task, to designing the skill, to writing the skill files, then validating that the wiring is correct, checking whether it actually triggers on good and bad prompts, and comparing the results against a baseline so I can see if it really improved. It should also include a careful review step that looks for scope mistakes, bad instructions, and weak evidence before I call anything finished.

Please make it work as a practical command line workflow, with clear checks for trigger tests, validation, and previewing evaluations before any live run. Keep the prompts and outputs measured and honest, since triggering can vary by model. If anything depends on current Claude Code behavior or docs, look up the latest details online and use those.