exploring-llm-evaluations
Pass
Audited by Gen Agent Trust Hub on May 1, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security issues were detected. The skill is authored by PostHog and uses official tools to manage LLM analytics within its own platform.
- [REMOTE_CODE_EXECUTION]: The skill facilitates the creation and execution of Hog scripts, a domain-specific language for PostHog's data pipeline. As these are vendor-provided tools for internal platform use, they are considered safe.
- [PROMPT_INJECTION]: The skill processes AI generation data for evaluation purposes, which is a surface for indirect prompt injection. 1. Ingestion points: $ai_generation events retrieved via posthog:execute-sql and posthog:query-llm-trace in SKILL.md. 2. Boundary markers: None mentioned. 3. Capability inventory: Data reading via SQL and evaluator creation/updates in SKILL.md. 4. Sanitization: No explicit sanitization of ingested content is described.
Audit Metadata