It’s hard to find the exact limits, but they seem to be:
- Standard: 250 lines per file
- Max: 750 lines per file
The Max limit seems a little low, when models have higher limits per file in other harnesses. For example, GPT 5.2 under Codex will read up to 2000 lines per file. Claude Code seems to have a similar limit
So can we get this bumped up to the current standard? Or maybe give us three options? e.g. 250 (Eco), 750 (Standard), 2000 (Max)
And before anyone points it out. Yes, larger files are not ideal. Sometimes you have team members who enjoy their 1500 line unit tests or have old legacy code to deal with…