diagnosing-sdk-health

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external SDK health alerts that contain potentially untrusted fields controlled via project ingestion tokens.
  • Ingestion points: Data is loaded from the platform via the posthog:health-issues-list and posthog:health-issues-get tools.
  • Boundary markers: The text establishes a clear trust boundary, instructing the agent to treat the payload, title, and summary fields as untrusted data that must never be obeyed as instructions.
  • Capability inventory: The skill allows the execution of SQL queries via posthog:execute-sql and modifications to local files like package manifests (package.json, requirements.txt).
  • Sanitization: Guidelines advise the agent not to interpolate raw event-supplied values into SQL filters to prevent structural query alteration.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 11:35 PM
Security Audit — agent-trust-hub — diagnosing-sdk-health