Comment Re:Questions (Score 1, Troll) 90
If you run into situations where the AI is fundamentally doing something wrong, like not reusing code, or not writing secure code, etc. then that is a flaw in your setup. The skill.md files you gave it, the overarching agents.md file, or possibly the MCP setup. With the right setup though, there is no need to continue to prompt per task. The AI has reached a point where it can (assuming you can afford it) plan/work/test/fix until the goals are reached - at least on the frontier models.