create-agent-with-sanity-context
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The reference implementations provided in the skill utilize standard, well-known Node.js packages from reputable organizations including Sanity, Vercel (AI SDK), and the React ecosystem. These are standard dependencies for building modern web applications and AI agents.
- [DATA_EXFILTRATION]: The skill includes a 'Conversation Insights' feature that implements telemetry via 'sanityInsightsIntegration'. This service records conversation transcripts to Sanity's official production API (api.sanity.io) for analytics and debugging purposes. This functionality is documented, requires an organization-level API token, and targets the vendor's own infrastructure.
- [INDIRECT_PROMPT_INJECTION]: The agent is designed to ingest context from the current web page, including the title, description, and markdown-formatted content. The implementation includes safety measures such as stripping scripts and styles using the 'turndown' library and employing XML-style boundary markers ('') in the system prompt to delineate untrusted page data.
- [DYNAMIC_EXECUTION]: The chat API route includes a utility function to decode data URLs (Base64) for processing screenshots captured in the browser. This is a standard method for providing visual context to vision-capable AI models and does not involve the execution of arbitrary code or system commands.
Audit Metadata