How it works
AI agents support the investigation. Zaaf researchers own the technical verification, the impact judgement and the approval of everything you receive.
Define · Investigate · Validate
Agree what the work is meant to answer.
Your context decides which systems and workflows belong in the Countermind.
Checkpoint
Testing begins only within the access and actions agreed in writing.
Trace
Explore hypotheses. Collect observations.
AI agents support research across the authorised workflows.
Checkpoint
Unknowns and testing limits stay explicit. They never become unsupported conclusions.
Trace
People approve what reaches your team.
A Zaaf researcher reproduces every finding, judges its impact and signs it off.
Checkpoint
The report carries reviewed findings, limits and remediation guidance. Nothing else.
Trace
04Delivery
Explain findings and limits
Illustrative
Every observation is reported in the state the evidence supports. Nothing is promoted to a finding because it looks likely.
| State | What it means | How it is reported | Produced in |
|---|---|---|---|
| Testing limitation | Access, environment or operational constraints prevented a particular check. | The report names the coverage gap instead of implying a clean result. | 01 Define |
| Open question | The available observations do not establish the proposed conclusion. | The uncertainty is named. It is never turned into a confirmed finding. | 02 Investigate |
| Simulated condition | A controlled test uses constructed or adjusted conditions. | The constructed conditions and the boundaries of the claim are stated. | 02 Investigate |
| Supported finding | The reviewed evidence supports a specific mechanism and an observed result. | Affected scope, preconditions, impact and relevant limits are explained. | 03 Validate |
Designated infrastructure supports approved workflows. Capability never replaces authorisation.
Records on our own domains and designated test servers support controlled scenarios.
Owned resourcesDedicated email accounts drive registration, verification and sign-in in agreed workflows.
Designated accountsAssigned wallets with small balances support approved transactions within defined environments and limits.
Explicit transaction scopePhysical-device testing is in development and is not offered today. Research resources are never unrestricted access to your systems.
A human approval gate only matters if the output helps your team understand the evidence and act on it.
See the reportWhat was included, excluded and constrained by the access available.
Mechanism, preconditions and the observations behind each conclusion.
Demonstrated consequences, kept apart from uncertainty and assumptions.
Remediation guidance and retest terms, where agreed.
They support investigation and evidence collection inside the agreed scope, including designated authenticated workflows and test infrastructure. Countermind is not an unreviewed autonomous scanner.
Researchers own the technical verification, the impact judgement and the approval of everything delivered to you.
Yes. Inconclusive observations and untestable areas are named. They are not proof that an issue exists, or that a system is secure.
Disruptive actions are never assumed to be in scope. Any such need is discussed explicitly: authorisation, operational impact and whether the work can support it.
A pentest is usually a tester's time-boxed sample. In a Countermind, AI agents work across the agreed scope and a named researcher reproduces and signs every finding before you see it.
No. Every Countermind is bounded by scope, access, environment and time. This page describes how the work is done, not a guarantee of completeness.
Thirty minutes with a Zaaf researcher. You bring one scope; we show you what a Countermind would look at first.