feature-usage-feed
Pass
Audited by Gen Agent Trust Hub on May 5, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: Identifies a potential indirect prompt injection surface. The workflow involves passing untrusted user-generated trace data to an LLM judge for classification and summary. The skill provides template-based boundary markers (headers and forced openers) to mitigate this, though runtime data content remains a risk factor inherent to the use case.
- [EXTERNAL_DOWNLOADS]: References official PostHog development documentation and pull requests on GitHub (github.com/PostHog/posthog). These are legitimate resources from the skill author used to demonstrate session linkage patterns.
Audit Metadata