[Bug] False positive cyber classification during legitimate local plugin development inspection
Status Open
Reported on v2.1.276
Maintainer reply None cached
Activity 0 comments · opened Sep 18, 2026
Bug Description
False positive: legitimate coding work flagged as [cyber], forced a model switch from Fable 5.1 to Opus 4.8 mid-task.
I was inspecting a locally-installed Herdr sidebar plugin (herdr-agent-quota) to add a feature: reading how it collects subscription quota, grepping the local CLI binaries for the endpoint/schema it uses, and checking my own credential file's structure (key names only, no secret values). This is ordinary plugin development on my own machine, not an attack. The [cyber] classifier misread it and switched models, interrupting the task.
Please tune the [cyber] safeguard so authorized local dev/config inspection (own credentials, own machine) isn't caught.
Environment Info
- Platform: linux
- Terminal: WezTerm
- Version: 2.1.276
- Feedback ID: e81f52fe-eb43-457b-a007-b67111317643
Errors
[]