joeseesun/qiaomu-goal-meta-skill ? reverse-engineered prompt

Reverse engineered prompt

Build me a small Python based agent skill that turns a vague task into a strong copy ready /goal command.

I want it to help people who only know how to say things like “make me an app” by turning that into something an AI coding agent can actually execute, verify, and know when to stop. It should generate a recommended version first, with outcome, verification, constraints, boundaries, iteration policy, completion evidence, and pause conditions, and it should keep the command prefix as /goal. If the task is risky or in a domain like medical, finance, copyright, or production data, it should lean toward discovery first and be careful about what it will not do.

Please make it easy to install and use from the command line, include a simple lint or check script for the generated goal text, and add clear docs in English and Chinese. If you need current tool or agent docs, look them up online first.