double-check

sagerstack 的头像作者 sagerstack

如果代理声明"工作已完成并可投入生产"但实际上通常不是,强制其再次思考的简单方法。此技能确保AI代理验证其任务,用户不必在代理工作完成后检查是否完成了工作。

📖 中文导读:如果代理声明"工作已完成并可投入生产"但实际上通常不是,强制其再次思考的简单方法。此技能确保AI代理验证其任务,用户不必在代理工作完成后检查是否完成了工作。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: double-check description: An easy way to force agent to think again if it's statement that the "Job is done and production ready" is actually done - usually it's not. This skill ensures that that the AI agent verifies their task and user does not have to check after the agent if they did their job.

When to use

  • After a task is completed and before reporting completion details to the user
  • After generating tests
  • After fixing tests

Skill

Please double check your work and what you generated.

  • Is it really complete?
  • Did you approach it from different angles?

Define first at what angles you can approach it. Reiterate on the goal and what it means and what is required to achieve it. Define what "complete" means in this context.

If the task was to generate tests or fix failing tests, run the tests to make sure they are successful. This is the only definition for "complete"

Proceed with the double check only after you are sure you have done all of the above.

Ultrathink!