conversion-carousel
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires access to an external Model Context Protocol (MCP) server located at
https://mcp.trylexsis.com/mcp. This server provides the necessary tools for storefront management, catalog access, and asset handling. - [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from project documentation, live web pages, and customer review libraries.
- Ingestion points: Project docs, brand context, product catalog, live storefront pages, and third-party review libraries (SKILL.md, Phase 1).
- Boundary markers: The instructions do not define explicit string delimiters for data isolation, but they enforce a logical separation through a four-phase workflow.
- Capability inventory: The skill uses tools to read storefront context and write to page sources and drafts (
lexsis_pages.edit_context,lexsis_pages.source,lexsis_drafts.page_update_section). - Sanitization: The skill implements a 'claim gate' which requires every fact to be verified against an evidence ledger before being used in output, specifically instructing the agent to ignore suspicious or unverified reviews.
Audit Metadata