build-zoom-virtual-agent
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSNO_CODE
Full Analysis
- Indirect Prompt Injection Surface: The documentation describes a Knowledge-Base Sync Pipeline for ingesting data from external content sources like CMS platforms and web pages. This pattern of processing untrusted external data represents a potential surface for indirect prompt injection.
- Ingestion points:
scenarios/high-level-scenarios.md(Knowledge-Base Sync Pipeline, custom API connector). - Boundary markers: The reference files do not explicitly specify delimiters or boundary markers for ingested content.
- Capability inventory: The reference architecture involves native bridge handlers (
exitHandler,commonHandler), support handoff mechanisms, and session management (SKILL.md,RUNBOOK.md). - Sanitization: The
scenarios/high-level-scenarios.mdfile notes policies for governing trusted versus external links, which provides a layer of validation for navigation outcomes. - External Resource Reference: The skill refers to official repositories for implementation samples. This is a standard approach for providing developer resources.
- Evidence:
references/samples-validation.mdincludes references to the officialzoomorganization repositories on GitHub.
Audit Metadata