mcp-inspector
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes output from external MCP servers to perform diagnostics, creating a surface for indirect prompt injection.\n
- Ingestion points: Untrusted data enters the agent context via the output of mcpjam commands (such as
server probe,tools list, andtools call) defined inSKILL.md.\n - Boundary markers: The skill instructs the agent to use structured JSON output for tool responses, but does not provide specific delimiters or ignore-instructions to isolate potentially malicious data within those payloads.\n
- Capability inventory: The agent can execute numerous
mcpjamsubcommands including network-facing probes, tool execution, and credential management as defined inSKILL.md.\n - Sanitization: No explicit instructions for sanitizing or escaping the received external data before it is used to drive the investigation logic are present.
Audit Metadata