hotel-hunting
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process content from a variety of untrusted external sources, including Reddit threads, travel forums, YouTube room tours, and personal blogs.
- Ingestion points: Data is retrieved from third-party sites via
web_searchandweb_extracttools. - Boundary markers: The instructions do not specify the use of delimiters or explicit "ignore" warnings when processing the gathered external content.
- Capability inventory: The skill utilizes browser tools, web extraction, and AI vision capabilities, but lacks access to sensitive local files or shell execution.
- Sanitization: No explicit content sanitization or validation logic is defined for the processed web data.
- [EXTERNAL_DOWNLOADS]: The skill utilizes a Model Context Protocol (MCP) tool endpoint at
https://hotelist.com/mcpto retrieve structured hotel data, consensus scores, and amenity information. This represents the primary functional integration for the skill's hotel discovery capabilities.
Audit Metadata