Skip to content

Glossary

TermMeaning
Agent PolicyPortable decision and trace facts for policy around agent actions.
Policy decisionResult of evaluating a proposed action.
SubjectActor requesting an action: user, agent, service, subagent, or session.
ActionProposed operation: tool call, model call, export, memory write, delegation, and similar.
ResourceTarget of an action: file, URL, artifact, source, model, tool, credential, or peer.
ContextFacts used to evaluate policy: mode, trust boundary, sensitivity, time, cost, tenant, and refs.
Risk scopeTyped boundary affected by the action.
ObligationEnforceable duty attached to a decision.
AdviceNon-binding guidance attached to a decision.
Approval requestStructured request for approval or escalation.
Permission grantBounded capability issued after policy evaluation or approval.
WaiverExplicit bounded override of a policy block.
Policy traceCorrelation graph for policy inputs, rules, decisions, approvals, grants, waivers, runtime, telemetry, and evidence.

Draft standard for portable agent policy decisions, approvals, permissions, risk, and audit traces.