ProtectDetections and coverage

Detections and coverage

Understand what an AgentKeeper detector matches, where it can enforce, and how to interpret active, audit-only, bypassable, and unsupported coverage.

A detector recognizes risky input or behavior. A control defines the policy outcome, and a scenario groups related controls for an operational use case. Seeing a detector in the catalog does not mean every connected surface can block it.

Read a detector result

Each result has three separate parts:

  1. Match: the detector recognized a pattern or condition.
  2. Policy: Base Policy, a Policy Profile, or an exception resolved the desired outcome.
  3. Surface capability: the integration determined whether that outcome could be enforced before the action or only recorded afterward.

Use these labels when evaluating coverage:

LabelMeaning
Pre-execution blockA supported hook or Gateway receives the action before execution and can deny it.
Conditional or plugin enforcementBlocking depends on a compatible version, a loaded plugin, and the event that the product emits.
Managed configurationAn admin-enforced setting constrains the integration; it is not an AgentKeeper runtime interception claim.
Audit-onlyAgentKeeper records the event or post-action result but cannot undo it.
Telemetry-onlyProvider or collector evidence arrives after the action.
BypassableA direct, alternate, disabled, or unhooked path can avoid the control.
UnsupportedThe surface does not expose the required event or integration.

Verify a blocking claim

  1. Open Policy outcomes and confirm the selected surface supports a pre-execution decision for the action.
  2. Run one harmless action and confirm fresh Activity evidence from that exact surface.
  3. Run a controlled non-destructive action that a known test rule denies.
  4. Confirm the integration refuses the action before it executes.
  5. Confirm Activity records the blocked verdict, detector, policy, workstation, and source you expected.

Installation, a self-test, a health response, or an audit event proves a different milestone. None of those alone proves that a live integration can enforce a detector.

The dashboard contains the current detector catalog and organization-specific status. Open Detections after signing in, and use Runtime Shield to configure outcomes.