II.
HumanCheckpoint overview
Reference · livehuman-checkpoint:dangerous-tool-approval
Dangerous-tool approval overview
Human approves a tool call flagged by the runtime as destructive (file delete, force-push, paid API). Hard block; matches Claude Code's permission-prompt UX.
Attributes
displayName
Dangerous-tool approval
kind
approval
blockingPolicy
hard
description
Human approves a tool call flagged by the runtime as destructive (file
delete, force-push, paid API). Hard block; matches Claude Code's
permission-prompt UX.
urgency
critical
interactionKind
sync
Outgoing edges
escalates_to2
- role:incident-commander·RoleIncident Commander
- role:security-engineer·RoleSecurity Engineer
has_strategy1
- breakpoint-strategy:multiple-choice·BreakpointStrategyMultiple Choice
notifies_via2
- channel:cursor-bg-agent·ChannelCursor background agent channel
- channel:claude-ai-web·ChannelClaude.ai web channel
routes_to_responder1
- responder-profile:on-call-incident-commander·ResponderProfileOn-call Incident Commander
Incoming edges
None.