azure-ai
Pass
Audited by Gen Agent Trust Hub on Nov 18, 2026
Risk Level: SAFE
Full Analysis
- Privacy-Preserving Diagnostic Tools: The skill includes several Node.js scripts (e.g.,
analyze-speech-sdk-log.mjs,inspect-document-input.mjs) designed to perform local analysis of user-provided files such as SDK logs and document inputs. These scripts use an allowlist approach to extract specific technical metadata and diagnostic observations rather than ingesting raw, potentially sensitive content directly into the agent's context. - Secure Authentication Guidance: The documentation (e.g.,
auth-best-practices.md) provides well-reasoned security advice for Azure developers, emphasizing the use of Managed Identities and Azure RBAC over hardcoded keys or the less deterministicDefaultAzureCredentialin production environments. - Input Validation: The provided validation scripts perform robust checks on file headers (magic bytes), sizes, and schemas before processing, which helps prevent issues related to malformed or unexpected input formats.
- Least Privilege and Safety Awareness: The skill's instructions explicitly remind the agent to avoid requesting credentials or echoing raw prompts/documents into the chat. It also contains guidance on treating logs as untrusted data to mitigate potential indirect prompt injection risks from log content.
Audit Metadata