hithink-finance-mcp
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill instructions direct the agent to access local credential files (e.g., 'credentials.env' in %APPDATA%, 'Library/Application Support', or '.config') and environment variables (e.g., 'HITHINK_FINANCE_API_KEY') to authenticate with the vendor's MCP endpoints. These credentials are used to populate the 'X-api-key' header for requests to the vendor domain at 'fuyao.aicubes.cn'.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from financial news, announcements, and reports, which may contain malicious instructions. Ingestion points: Untrusted content enters the agent's context through tools such as 'get_news_article_detail' (HTML body) and 'get_news_notice_detail' (announcement text). Boundary markers: The skill lacks explicit delimiters for processed data, although instructions mention following a 'content safety policy' and returning only summaries. Capability inventory: The skill possesses network access via high-level MCP data tools and the capability to write large result sets to the local file system ('落盘'). Sanitization: The documentation instructs the agent to apply data masking ('脱敏') and follow content safety protocols when displaying retrieved article content.
Audit Metadata