OptionalallowDANGEROUS ESCAPE HATCH. When true, allows invocation of ANY action via
utilities.actions.Invoke, ignoring allowedActions. Only set for
framework-authored orchestrators. Approval UI flags this.
OptionalallowDANGEROUS ESCAPE HATCH. When true, allows invocation of ANY agent via
utilities.agents.Run, ignoring allowedAgents. Only set for
framework-authored orchestrators. Approval UI flags this.
OptionalallowDANGEROUS ESCAPE HATCH. When true, allows access to ANY entity via
utilities.md.*, utilities.rv.*, and utilities.entity.*, ignoring
allowedEntities. Only set for framework-authored utility actions that
accept the target entity as runtime input. Approval UI flags this.
Other actions this Runtime action can invoke via utilities.actions.Invoke
Agents this Runtime action can run via utilities.agents.Run
Entities this Runtime action can read or mutate via utilities.rv / utilities.entity
Declarative permission scopes for a Runtime action. The bridge enforces each scope on every call — an attempt to touch an unlisted entity / action / agent throws a PermissionDenied error before the downstream operation runs.
IDs are the source of truth; names are kept alongside for display, logging, and human review during the approval workflow.
The
allowAnyEntity/allowAnyAction/allowAnyAgentbooleans are escape hatches for framework-shipped utility actions that must accept the target entity/action/agent as runtime input (e.g. a generic "data quality report" that can analyze any entity). They bypass the allowlist entirely for their namespace. The approval UI renders a prominent warning when any of them is set so a human reviewer sees the blast radius at approval time; agent-authored Runtime actions should enumerate specific references rather than set these flags.