Skills
Navigate
skills/posthog/skills/debugging-mcp-analytics/Gen Agent Trust Hub

debugging-mcp-analytics

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill describes a system that summarizes agent-generated intent strings, creating an indirect prompt injection surface. The documentation notes that these risks are addressed through built-in SDK sanitization and data structure design.
  • Ingestion points: The $mcp_intent property and captured tool parameters/responses as described in SKILL.md and references/event-vocabulary.md.
  • Boundary markers: The skill explains that conversation handles were changed to plain JSON to prevent agents from interpreting them as instructions (references/stateless-and-sessions.md).
  • Capability inventory: The documented capabilities include ClickHouse query runners and Temporal workflows for background data processing.
  • Sanitization: Recent SDK versions automatically redact exception messages and large binary payloads (references/event-vocabulary.md).
  • [EXTERNAL_DOWNLOADS]: The documentation references official PostHog packages including @posthog/mcp, @posthog/wizard, and the posthog Python library.
  • [COMMAND_EXECUTION]: The skill provides examples of using the vendor's CLI via npx -y @posthog/wizard@latest mcp-analytics to instrument servers.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 10:55 AM
Security Audit — agent-trust-hub — debugging-mcp-analytics