Fixed scope
$250
One-time review of one agent workflow and its declared integrations. Delivery target: 7 days after scope acceptance.
Persistent and connected business agents
Prove what a long-running agent may do after the prompt window closes.
A fixed-scope evidence audit for organizations deploying event-driven, headless, connected, or multi-agent systems that can keep working across channels and systems of record.
Authority chain
- Principal, team, role, agent, and temporary sub-agent identities remain distinct.
- Delegated scopes, inherited permissions, and human approvals are explicit.
- Events and schedules cannot silently expand authority.
Runtime evidence
- Persistent jobs expose checkpoints, expiry, cancellation, and completion receipts.
- Connectors, system-of-record writes, sandboxing, and network routes are bounded.
- Memory retention, source lineage, model routing, spend caps, and exception paths are inspectable.
- Cross-domain actions preserve intent, delegation, authorization changes, actor identity, service-side receipts, and causal order.
Deliverable
- Actor-to-effect authority map.
- Trigger, connector, memory, spend, and audit-context evidence register.
- Failure, cancellation, and rollback gaps.
- Prioritized remediation and one bounded retest.
Why now. Google introduced the Gemini agent on October 8, 2026 as a persistent, event-responsive, headless agent that can connect to business systems and dynamically create temporary sub-agents. Read Google Cloud's primary announcement.
An IETF Internet-Draft proposes linking user intent, evolving authorization, agent and sub-agent delegation, service-observed actions, attestations, and transparency receipts across administrative domains. It is work in progress—not a standard—but it exposes the evidence gap between isolated logs and a reconstructable causal chain.
The audit tests the operational evidence those capabilities require; it is not a Google certification, IETF conformance claim, or product endorsement.