exploring-mcp-tool-usage
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted external data retrieved from Model Context Protocol (MCP) event logs, which constitutes an indirect prompt injection surface.
- Ingestion points: The skill retrieves agent intents and error messages from
$mcp_tool_calland$exceptionevents using tools such asposthog:query-mcp-tool-sample-intentsandposthog:query-mcp-tool-failures. - Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the retrieved log strings.
- Capability inventory: The skill has access to PostHog analytics tools and the
posthog:execute-sqlcapability. - Sanitization: No explicit sanitization or filtering logic is described for the retrieved intent or error strings.
Audit Metadata