On Cursor Version: 0.49.4 (Universal), MacOS Sequoia 15.3.2
This has never worked for me-- I’ve always had to highlight all the terminal output and copied as HTML in order for the agent to read the terminal output. Clicking ‘Copy’ does the same thing as clicking ‘Add to Chat’, in that it shows the terminal lines in the chat context, but the terminal output isn’t passed to the agent.
Here is an example where I want the agent to read the terminal output of a series of e2e tests, and try to fix the failing tests, and it doesn’t see the terminal output, so it instead assumes I’m asking about linter errors.
-
Run tests in terminal & click ‘Add to Chat’
-
With context loaded, prompt the agent to solve the errors
-
Agent thinks I’m asking about linter errors because it can’t see the terminal output