| name | human-management |
| description | Use when managing human users — creating accounts, setting permissions, managing access to teams and workers. |
Human Management
Manage human user accounts and their access to the AgentHub platform.
Commands
agenthub schema
agenthub human create --name <username> --display-name "John Doe" --email john@example.com --permission-level 1
agenthub apply -f human.yaml
agenthub apply -f human.yaml --approval-mode request --room <room-id> --reason "Human permission change"
agenthub approval confirm --event <approval-timeline-event-id> --approved-by <human-id> --reason "Human approved in room"
agenthub approval deny --event <approval-timeline-event-id> --denied-by <human-id> --reason "Human denied in room"
agenthub human list
agenthub human delete --name <username>
Human Manifest
kind: Human
metadata:
name: admin-user
spec:
displayName: Admin User
email: admin@example.test
permissionLevel: 1
Permission Levels
- Admin (Level 1): Full access — can manage workers, teams, and platform settings.
- Team (Level 2): Can interact with assigned teams and their workers.
- Worker (Level 3): Can observe and interact with assigned workers only.
Rules
- Human users are first-class collaborators, not just observers.
- New human users should be announced in the relevant room.
- Permission changes take effect immediately.
- Human DM access to the Manager is controlled by the DM allowlist in openclaw.json.
Decision Pattern
agenthub schema to inspect humans.create and approval metadata.
- Read the room timeline and determine what human action or permission change is needed.
- For sensitive changes, write a Room-native approval request with
agenthub apply -f <manifest> --approval-mode request --room <room-id> instead of applying immediately.
- When the human approves or denies, resolve the request with
agenthub approval confirm or agenthub approval deny.
- Apply the human account, room invite, or permission update through Controller APIs only after confirmation when required.
- Announce the result in the room.
- Treat human clarification and approval messages as authoritative task context.