azure-fluid-relay
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is configured to retrieve and process documentation from external sources, specifically learn.microsoft.com. This behavior introduces a surface for indirect prompt injection where adversarial instructions could be embedded in the documentation content.
- Ingestion points: Documentation URLs targeting learn.microsoft.com using fetch_webpage or mcp_microsoftdocs tools.
- Boundary markers: The skill does not provide specific delimiters or instructions to the agent to treat the fetched content as potentially untrusted data or to ignore embedded instructions.
- Capability inventory: The skill uses tools for network-based documentation retrieval; it does not request high-risk capabilities such as arbitrary shell execution, persistent storage modification, or system-level configuration changes in the SKILL.md.
- Sanitization: The skill lacks explicit sanitization steps or validation logic for the content fetched from the remote documentation service.
- [EXTERNAL_DOWNLOADS]: Fetches documentation from learn.microsoft.com and references an official installation guide for a documentation tool on the MicrosoftDocs GitHub repository.
Audit Metadata