Update prompt.txt

[sneljeroen]

Sometimes it tries to execute commands like "Google Search" instead of "google", specifying clearly which one is the command name fixes this.

[PhilipAD]

lgtm

[Void-n-Null]

Thats a lot of extra tokens, but if its effective in avoiding miscommunication on command names, it might be worth it.

[fire17]

i would suggest #172 here:
https://github.com/Torantulino/Auto-GPT/pull/172/files

[mis046]

If token count is a concern, structure the data as a table might help e.g.
instead of

1. Google Search; command name: "google", args: "input": "<search>"
2. Memory Add; command name: "memory_add", args: "string": "<string>"

might be able to use

command key | command name | args
Google Search | "google" | "input": "<search>"
Memory Add | "memory_add" | "string": "<string>"

Tried running this prompt

{
  "prompt": "Here is a table with | as the column delimiter and new line as row delimiter, with the first row being the header:\ncommand key|command name|args\nGoogle Search|\"google\"|\"input\": \"<search>\"\nMemory Add|\"memory_add\"|\"string\": \"<string>\", what is the value of row \"Memory Add\"'s command name field",
  ...
}

and this is the result

{
  "choices": [
    {
      "text": "?\n\n\"memory_add\"",
      "index": 0,
      "finish_reason": "stop",
      "logprobs": null
    }
  ], 
  ...
}

[nponeccop]

1. Make the PR title more descriptive
2. Resolve the conflicts (preferrably by rebasing on top of current master)

[github-actions]

This pull request has conflicts with the base branch, please resolve those so we can evaluate the pull request.

[k-boikov]

This PR is completely obsolete.

[BillSchumacher]

Yeah this doesn't even exist with plugins, sorry.