stock-announcement-analysis
Fail
Audited by Gen Agent Trust Hub on Jun 25, 2026
Risk Level: HIGHCREDENTIALS_UNSAFEDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill hardcodes a Tushare Pro API token ('c8dbb3833192a3e47991b1975ad02d95a6567988826e519ba76b0ef5') in a Python script snippet.
- [DATA_EXFILTRATION]: The code directs the Tushare library to use a non-standard HTTP URL ('http://tushare.xyz') instead of the official API endpoint, which could lead to credential interception.
- [COMMAND_EXECUTION]: Uses shell commands ('curl') to interact with external financial data endpoints.
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. * Ingestion points: Reads content from external PDF announcements (CNINFO) and user-provided Excel files. * Boundary markers: No clear delimiters are used to separate ingested content from system instructions. * Capability inventory: Access to network APIs (requests, browser tool) and the ability to process and output structured data. * Sanitization: No validation or sanitization of the PDF or Excel content is performed before analysis.
Recommendations
- AI detected serious security threats
Audit Metadata