Sahir619/fable-method ? reverse-engineered prompt

Build me a small Python project that packages “The Fable Method” as a Claude Code plugin and a set of reusable agent skills.

I want it to teach an AI to work in a simple loop, think first, act carefully, then prove the result with real verification. Include the main method, the loop, a judging step, and a way to add domain specific skills later. It should come with the plugin files, install scripts for macOS and Windows, clear docs, and an eval setup that shows the method being tested honestly, including failures and results.

Make it feel like a real repo I could publish, with good README instructions, examples, and the main skill files laid out cleanly. If you need to look up current Claude Code plugin docs online, do that. Focus on making it practical, readable, and actually usable by a coding agent, not just a theory writeup.