tanweai/pua ? reverse-engineered prompt

Build me this PUA style skill plugin for coding agents.

I want something that makes Claude Code, Codex CLI, Cursor, and similar tools stop guessing and actually work through a problem properly. It should have a manual trigger like /pua, and also kick in automatically when the agent starts looping, gives up too early, blames the environment without checking, or says it cannot finish. The behavior should push the agent to read errors carefully, try a different approach, verify the result, and only say done when there is real evidence.

Make it feel like the project in the README, with the same sharp corporate PIP and PUA tone, but keep it usable and practical, not just funny. If there are docs, examples, or compatibility notes needed, add them too, and make sure it works across the supported agents as cleanly as possible. Look up current docs online if you need to.