skills/unbrowse-ai/sites/wikimedia/Gen Agent Trust Hub

wikimedia

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill connects to a remote Model Context Protocol (MCP) server at https://unbrowse.ai/mcp/wikimedia.org. This is a vendor-owned resource used to provide tool functionality.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes content fetched from wikimedia.org. 1. Ingestion points: Output from the commons_wikimedia_org__render_page tool. 2. Boundary markers: None defined for external content. 3. Capability inventory: The skill provides access to cloud-based browser and network tools via the unbrowse namespace. 4. Sanitization: No explicit sanitization is described in the skill instructions.
  • [DYNAMIC_EXECUTION]: The skill documentation mentions that using browser automation tools can result in the dynamic creation of new tools on the remote server. This process is managed by the vendor's cloud infrastructure.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 04:13 AM