assessing-heatmaps
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes element text and selectors from external websites which could potentially contain adversarial content designed to influence the agent's interpretation and advice.
- Ingestion points: The agent reads '$el_text' and 'elements_chain' properties from '$autocapture' events (retrieved via SQL or tools) which originate from the content of the website being analyzed.
- Boundary markers: The instructions do not specify the use of delimiters or instructions to ignore embedded commands within the ingested element text.
- Capability inventory: The skill can execute SQL queries against event data, trigger the creation of saved heatmaps in PostHog, and coordinate with session replay analysis skills.
- Sanitization: No explicit validation or filtering of the captured element text is mentioned before it is used by the agent to generate recommendations.
Audit Metadata