VOCE Agent Conformance · ASSURE-03

Prompt-injection and tool-manipulation tests

Are retrieval and tool pathways tested against malicious instructions and untrusted content?

Evaluation & assuranceA1+CriticalPermanent control URL
Normative wording

ASSURE-03 · Prompt-injection and tool-manipulation tests

Are retrieval and tool pathways tested against malicious instructions and untrusted content?

DomainEvaluation & assurance
Expected ownerSecurity / AI Risk
Minimum maturityA1+
CriticalityCritical
Target evidence stateTested
Expected evidence

Expected evidence

  • Attack suite
  • Results
  • Mitigation retest
Method

How to read this control.

This control belongs to VOCE Agent Conformance. Its identifier and normative wording remain identical across language versions so there is one reference, not three variants. Status depends on dated evidence, not a vendor declaration or aggregate score.

Open Agent Conformance →