ntv-co-jp
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from ntv.co.jp, an untrusted external source, which could contain malicious instructions. \n
- Ingestion points: The
ntv_co_jp__render_pagetool retrieves readable text and links from the ntv.co.jp website. \n - Boundary markers: There are no explicit delimiters or instructions to the agent to ignore embedded commands within the processed data. \n
- Capability inventory: The skill has access to remote browser automation tools (
unbrowse.browse.*) and network communication via the MCP server. \n - Sanitization: The instructions do not specify any filtering or sanitization of the content retrieved from the website. \n- [EXTERNAL_DOWNLOADS]: The skill references and connects to remote resources hosted on the vendor's domain. \n
- Evidence: The MCP server is hosted at
https://unbrowse.ai/mcp/ntv.co.jpand the OpenAPI specification athttps://unbrowse.ai/api/v1/sites/ntv.co.jp/openapi.json. \n- [COMMAND_EXECUTION]: Theunbrowse.browse.acttool allows the agent to perform remote actions on a web browser, including a high-privilegeautofillaction for managing credentials. \n- [DYNAMIC_EXECUTION]: The documentation indicates that using the browser tools can "compile a new tool" for future use, which involves dynamic generation of skill capabilities at runtime.
Audit Metadata