exploring-live-traffic
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest user-supplied data (such as hostnames or bot names) and interpolate them into JSON query templates for product analytics. While this is the intended functionality, it creates a surface for potential schema confusion or indirect injection if the agent processes malicious instructions embedded within the user's input strings.
- Ingestion points: User-provided hostnames, bot names, and categories described in the "Common requests" and "Building product-analytics queries" sections of
SKILL.md. - Boundary markers: The skill uses structured JSON templates to define boundaries, but lacks explicit "ignore embedded instructions" warnings for the agent when handling user-provided strings.
- Capability inventory: The skill allows the agent to generate
TrendsQueryinsights which are then processed by the PostHog platform. - Sanitization: The instructions recommend stripping protocols from host strings (e.g.,
example.cominstead ofhttps://example.com), which provides a basic level of input normalization.
Audit Metadata