mcp-integration

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill documents how to integrate external data sources (APIs, databases, hosted services) which introduces an inherent attack surface where untrusted data could influence agent behavior.
  • Ingestion points: Data fetched from external services like Asana, GitHub, or custom database backends via MCP tools.
  • Boundary markers: The documentation suggests using the allowed-tools frontmatter to restrict agent access to specific tools, reducing the exposure surface.
  • Capability inventory: The integration patterns describe capabilities for filesystem access, network requests, and local process execution.
  • Sanitization: The guide emphasizes security best practices, such as avoiding hardcoded tokens and using environment variables, though specific data-scrubbing libraries for tool output are not mandated.
  • [EXTERNAL_DOWNLOADS]: Documents the use of npx to fetch and run official packages like @modelcontextprotocol/server-filesystem from the npm registry.
  • [COMMAND_EXECUTION]: Provides instructions for configuring local executable processes to function as stdio-based MCP servers.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 01:19 PM
Security Audit — agent-trust-hub — mcp-integration