Team Health

Open Settings > Security and data > Team or Projects > Activity. Healthy means no current issue crossed a threshold, Needs attention means work should be reviewed, and Critical means a credential, approval, active eval, trigger, budget, or repeated failure needs prompt action.

• Connections are checked against the exact capabilities granted to the workspace, including expired or revoked credentials.
• Approvals become stale after 24 hours; three expired windows on one automation in seven days become a critical pattern.
• Active and candidate golden eval results are tracked separately so a failing candidate stays out of traffic.
• Failures, completion rate, cost drift, hosted budget, and Trigger Hub health reuse bounded server-authoritative records shown elsewhere.

Most recommendations are navigation actions: reconnect in Apps, decide in Notifications, inspect Trigger Hub, review Activity, or check usage. Team Health never changes a model, permission, App grant, Team structure, budget, or active version automatically.

Manual repair still requires PAUSE AUTOMATION after a fresh server recomputation. Owners and admins may also explicitly enable Pause unhealthy automations. That policy pauses only an active automation with three consecutive failures, an active-version eval regression, or repeated approval timeouts; writes a version event and redacted audit; and emails the owner one recovery link.

Publishing workspace automation enrolls the workspace in the existing ten-minute hosted maintenance tick. Checks run in small batches; a failed check is deferred so it cannot monopolize the queue, and it cannot block trigger delivery or hosted runs. Snapshots contain safe issue summaries and aggregate counts, not logs, prompts, output, credentials, source data, or account identifiers.