[Bug] Overly aggressive safeguard flags during legitimate software development tasks
Bug Description
I'm frequently getting safeguard flags while working on my own software project, even though I'm only writing, debugging, and testing code within my own codebase. There is no cybersecurity, malware, biology, or other sensitive activity involved.
The safeguards appear to be triggering on routine software engineering tasks, which interrupts my workflow by switching models unnecessarily. This happens often enough that it's becoming difficult to use Claude for day-to-day development.
It would be helpful if the safeguards could better distinguish between legitimate software development on a user's own project and potentially harmful requests. If possible, please review the detection thresholds or provide more context about what specifically is triggering these flags so I can avoid false positives.
Thank you.
Environment Info
- Platform: darwin
- Terminal: vscode
- Version: 2.1.205
- Feedback ID: 66085e70-744d-45b7-9ce1-8674ee8204f7
Errors
[]This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗