[Bug] Safety filter false positive on internal tool-call messages during code analysis
Status Open
Reported on v2.1.202
Maintainer reply None cached
Activity 0 comments · opened Jul 18, 2026
Bug Description
False positive. I'm building a knowledge graph of our own legacy PHP monolith: a script parses source files, extracts state transitions, and loads them into Neo4j. Routine internal code analysis on a codebase we own — no security, exploit, or bio context anywhere in the session. The flag fired mid-task on an internal tool-call message, not on my prompt, and silently switched me to Opus 4.8.
Environment Info
- Platform: darwin
- Terminal: Apple_Terminal
- Version: 2.1.202
- Feedback ID: f3466856-8db1-41e1-8f8e-d8d57b00ed59
Errors
[]