anthropics / anthropics/claude-code
[Bug] Model unexpectedly switched during coding session
- Vorherrschende Sprache
- Python
- Sterne
- 145k
- Forks
- 23.1k
- PR-Merge-Kennzahlen
- PR-Kennzahlen ausstehend
Beschreibung
**Bug Description**
Model switched mid-session possibly because we were investigating mechanisms to catch bad actors on our system(some SSH related work). Not ideal. Could have stopped and prompted me why its pausing and I could either understand and explain or open a new session.Credential hygiene checks are ordinary defensive security tasks. If the intentionally broad safeguard is going to catch them, the recovery path should be a prompt, not an unrequested downgrade.
**Environment Info**
- Platform: linux
- Terminal: vte-based
- Version: 2.1.266
- Feedback ID: 82ecbced-79b9-47ab-b2a0-70460c2337af
**Errors**
```json
[Fable 5.1'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]
```
Beitragsleitfaden
Für dieses Repository ist kein Beitragsleitfaden indexiert
Rechercherichtung
No source file, test, or entry point is identified. Start by reproducing the model switch in the reported Linux, VTE-based terminal environment using a defensive SSH-related task, then inspect the /config model-switch behavior. Done means legitimate security work no longer causes an unrequested switch and instead prompts for confirmation or explanation.
Vom Indexierungsmodell aus dem Issue-Text verfasst.
Bewertung
- Tech-Stack
- linux
- Bereich
- cli, security
- Issue-Typ
- Bug
- Schwierigkeit
- 5/5
- Geschätzter Aufwand
- Über eine Woche
- Aktivitätsstatus
- Aktiv
- Klarheit
- Größtenteils klar
- Anfängerfreundlichkeit
- 35/100