azure-ai-contentsafety-ts
Pass
Audited by Gen Agent Trust Hub on Mar 1, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill utilizes official Microsoft Azure libraries (@azure-rest/ai-content-safety, @azure/identity, and @azure/core-auth). All network communications are directed to trusted Azure Cognitive Services endpoints.
- [PROMPT_INJECTION]: The skill is designed to ingest and moderate external data, which creates a surface for indirect prompt injection attacks.
- Ingestion points: Untrusted data enters via the
textfield in the/text:analyzeoperation and theimage(base64 or blobUrl) field in the/image:analyzeoperation as documented inSKILL.md. - Boundary markers: The code samples do not demonstrate the use of delimiters or specific instructions to prevent the agent from inadvertently executing instructions contained within the text or image metadata being analyzed.
- Capability inventory: The skill includes
fs.readFileSynccalls to access local image files for the purpose of moderation analysis. - Sanitization: No client-side input validation or sanitization of the content to be analyzed is demonstrated in the skill's logic.
Audit Metadata