carousell-sg
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill represents an indirect prompt injection surface because it fetches and processes external content from Carousell listings.
- Ingestion points: Tools like
carousell_sg__get_search_macbook_airandcarousell_sg__read_pageingest data from carousell.sg. - Boundary markers: No specific boundary markers or delimiters for untrusted content are documented.
- Capability inventory: The skill possesses capabilities to search, read, scrape, and perform browser-based actions via
unbrowse.browse.actandunbrowse.run. - Sanitization: No sanitization or filtering of external data is specified before processing by the agent.
- [EXTERNAL_DOWNLOADS]: The skill documentation references the vendor's official domain (
unbrowse.ai) for connecting to the MCP server and fetching skill configuration.
Audit Metadata