mcp-builder-ms
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill directs the agent to fetch protocol specifications and SDK documentation from
modelcontextprotocol.ioand themodelcontextprotocolorganization on GitHub. These are recognized as the authoritative and well-known sources for the Model Context Protocol. - [INDIRECT_PROMPT_INJECTION]: The guide describes the creation of servers that ingest data from external APIs and expose it to an LLM, creating a potential surface for indirect prompt injection if the external data contains malicious instructions.
- Ingestion points: Data retrieved from third-party APIs and services via the custom MCP server (SKILL.md).
- Boundary markers: The guide mentions 'request-origin protections' and 'redacted errors' but does not include explicit instructions to use delimiters or ignore instructions within the ingested data content.
- Capability inventory: Servers built following this guide are intended to perform network operations and interact with cloud services like Azure, GitHub, and various databases.
- Sanitization: The skill recommends implementing input/output limits and authentication to manage and secure the data flow between the LLM and external services.
Audit Metadata