Persistent context and cancelling tools

[marcinmantke]

Hi!
Recently I started using this plugin extensively and I thought of two things that are interrupting my flow most often.
I'm mostly using chat with Claude 4 which has @copilot and @neovim plugins enabled by default. For most of the time, the suggestions I'm getting from the Copilot are pretty good, but sometimes I want to reject the change or even change something myself. And here comes the first problem - if Copilot tried to use a tool, there's no reliable way for me to make it drop the idea. So it's stuck in a loop trying to invoke a tool which I don't want. Is there any way to make it stop, other that resetting the chat?

And second question, because for now resetting the chat is the only reliable solution to the first problem. As I'm working on a more complex problem, I've saved a description of that problem to a file. Every time I reset the chat, I paste the content of the file to the first prompt, so the tools knows what I need to achieve and how do I plan to do it. Would it be possible to somehow attach the file instead of copy-pasting it to the prompt, something similar as Cursor is doing with context files?

[deathbeam]

Hmm so for the first problem, removing the tool from the prompt is not enough after LLM tries to call it? As then User skipped this function call is sent as tool response which should theoretically be able to make LLM stop calling it.

For second question, you can always just do #file sticky i guess and it should work, or just make custom function that provides some project-local markdown file.