Today I encountered issue multiple times while the agent was executing tasks, even after reverting to re-execute. This led me no clue on what the agents are doing, which made it hard to trace back activities. Additionally, there were several messages, such as “Model provided invalid arguments to read file tool." While I’m not sure if this is a direct consequence of the error, the result provided from the agent seemed not as good.
Please add the following info to help us diagnose your issue:
Saw this has been posted before but didn’t see an update on solution The agent just printed and “Model provided invalid arguments to read file tool." while executing the work Input prompt, choose Claude Sonnet 3.7 thinking, hit Enter Attached are screenshots.
Absolutely I’ve been working on the same project for about 2 weeks now, I would say about 30% of my day look at the other screen is what you showed on your screenshots.
I’m thinking it might be due to lag or the maximum amount of files, I tried restarting Cursor and starting new chats, but its very anecdotal evidence here I’m really not sure.
Furthermore, I don’t know if it does it to you, but after the 25 messages, more often than not, the “resume” button won’t work, gives you a timeout error. I don’t know if its related, but usually happens all around the same time.
for more context: it seems the issue only happened to Claude 3.7 both with & without thinking. however it didn’t happen with other models, ie: Claude 3.5 or Gemini 2.5, whose result isn’t as good compared to Claude 3.7 when it properly functions.
@dcastl now cursor just hid all the messages related to changes instead of showing < no message > - This is still challenging for me to understand what it’s changed. regarding “Model provided invalid arguments to read file tool error”, same as before. The result generated didn’t seem to follow through my prompt - perhaps you can look into it & give some suggestions?
Seeing some issues with our read tool prompt in your example, looking into it!
As for “cursor just hid all the messages related to changes”, in general we never expect to see < no message > when the model decides not to talk, we should just not see anything! So the “no message” behavior is actually pretty strange and would be great to get a repro.
attached screenshot & ID request today: Request ID: e4b1471c-d254-4f28-a6f9-63165f699c4c (there were 2 other issues happened in this request- 1) cursor got stuck in terminal command for hours and 2) cursor stopped running the request after I skipped terminal command)
Saw this has been posted before but didn’t see an update on solution
The agent just printed and “Model provided invalid arguments to read file tool." while executing the work
Input prompt, choose Claude Sonnet 3.7 thinking, hit Enter
Attached are screenshots.
OS sonoma 14.1 and your Cursor version 0.49.6
Probably…
