01 / The operating case
What changes in a real workflow?
An isolation agent received a valid grant for endpoint E-42; the incident is reassigned before the agent invokes the isolation tool.
- 01Authority origin
- 02Task and grant
- 03Current decision
- 04Observed outcome
02 / Decision boundary
Put the check where it can stop the effect.
At the isolation boundary, read current assignment and grant state; deny or escalate if the agent no longer owns the incident.
Architecture review / negative test
Make the boundary testable.
Identify which signals are authoritative and how fresh each must be. Incident ownership may come from a ticket system, while resource state comes from the target API. If either is delayed, define whether the operation pauses, escalates or narrows to read-only access. An untrusted prompt string is not an authoritative risk signal.
03 / Failure and evidence
What goes wrong, and what can be proven?
A decision cached at task start silently survives a change in operating context.
Record which time-stamped signals influenced the decision and which were unavailable.
04 / Canonical scope
Why this reference stands alone
Runtime caching examines freshness mechanics; this page owns the policy meaning of changing context.
Primary references