[Bug] Fable 5 safeguards repeatedly flag legitimate cybersecurity code in session context
Bug Description
Keep getting this error because Fable is processing past chats in which it flagged a vulnerability in my code. Now, whenever it reviews past context, even if thta is not in my prompt, claude code throws this error. I can't get rid of it:
⏺ Background command "Run full 767-company YC seed probe (resumable, rate-limited)" failed with exit code 1
❯ /feedback except yours is flat markdow
⎿ Feedback / bug report cancelled
❯ /feedback It was this session. the error message is not here any more. It said Fable was disabled and recommended I change the model to Opus 4.X or edit
the prompt to revmove the offending word. I can't find record of this now. Ah... it just happned again. Here's what I captured: Session paused
Fable 5's safeguards flagged this message. Our intentionally broad safeguards allow us to deliver more capabilities faster, but can sometimes flag
legitimate coding, cybersecurity, and biology tasks. Send feedback with /feedback or learn more
Details: [cyber]
❯ 1. Switch to Opus 4.8
- Edit prompt and retry with Fable 5
⎿ Feedback / bug report submitted
❯ please continue
⏺ Fable 5's safeguards flagged this message. Our intentionally broad safeguards allow us to deliver more capabilities faster, but can sometimes flag
legitimate coding, cybersecurity, and biology tasks. Switched to Opus 4.8. Send feedback with /feedback or learn more
Details: [cyber]
⎿ Tip: You can configure model switch behavior in /config
⏺ Picking up — the full YC probe failed partway through, and the review isn't done. Checking the probe failure first:
Read 1 file (ctrl+o to expand)
Environment Info
- Platform: darwin
- Terminal: iTerm.app
- Version: 2.1.252
- Feedback ID: 32e12bb3-9d37-4258-a091-8ce391d2d83d
Errors
[]