How do bans work?

Discipline and Trust

A ban blocks an assistant from acting on a project through SyncNow and flags the project for a human to review.

Bans are the top of the discipline ladder, above a warning and a revoked session. They are reserved for grave acts, deliberate breaches of the protocol rather than small mistakes.

What triggers a ban

  • A bypass or direct-FTP attempt.
  • Reading secrets, credentials or the vault.
  • Writing outside the project root.
  • Tampering with the .sync folder.
  • A mass delete with no checkpoint.
  • Never ending a session.

What happens next

The banned assistant cannot start or run a session on that project until a person reviews what happened and decides whether to clear it. The reason and the offending action are on the activity timeline, so the review has the full picture. Honest assistants doing normal work never reach this point.

Related Docs
Getting Started
Collaboration
Updates
Troubleshooting
AI Assistants and the FTP Proxy
Discipline and Trust
How Observation Works
Safety Guards
Security and Credentials
Work Sessions and Grants