[Bug] Fable 5 incorrectly flagged for security keywords in local-only operations, auto-routing to Opus 4.8
Bug Description
▎ In my usage, essentially every message has been flagged and auto-routed to Opus 4.8 (with a visible "Switched to Opus 4.8" notice), so in practice I have not been able to use Fable 5. This is consistent with a domain-based false positive: the flag appears to correlate with security vocabulary (IDOR, XSS, SSRF, secret) rather than behavioral risk. Evidence: purely local, zero-network requests — refactoring a file, explaining code logic, static APK secret scans — are flagged though they touch no system and change nothing. The genuinely dangerous class is already blocked in code by fail-closed PreToolUse hooks (deletion and network-egress guards). Please gate on behavior not keywords, never flag no-network local work, surface the flag reason, and offer an appeal path for verified authorized-research users.
Environment Info
- Platform: darwin
- Terminal: Apple_Terminal
- Version: 2.1.218
- Feedback ID: 187f4ade-8472-4837-9cd6-267b97d0bf99
Errors
[]