As an AI language model, I cannot ...

[Boostrix]

Duplicates

• [X] I have searched the existing issues

Summary 💡

I suppose that receiving "As an AI language model, I cannot [create files on your computer.]" is a sure sign that the generated prompt needs to be rewritten internally, so maybe we need some heuristics in place to detect this being a part of the response, to rewrite the prompt and get rid of it (it's wasteful and not useful in the Auto-GPT context)

I keep seeing these when using (many) sub-agents that prompt each other to do something via the inter-agent messaging API.

At the very least, such combinations of prompt/response should probably be logged separately, so that people can provide these back to the project, to hopefully improve the prompt generator over time ?

Examples 🌈

No response

Motivation 🔦

No response