cn-market-mechanics
Warn
Audited by Gen Agent Trust Hub on May 8, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill contains a hardcoded API key value (
123456) within theAUTHshell variable, which is used for all authenticated API requests. - [COMMAND_EXECUTION]: The skill relies on
curlto perform network operations, which involves executing shell commands on the host system. - [EXTERNAL_DOWNLOADS]: The skill communicates with an external service hosted on a raw IP address (
43.167.234.49). Using raw IP addresses instead of registered domain names makes it harder to verify the identity of the data source. - [DATA_EXFILTRATION]: User-provided parameters such as stock codes (
tsCode) and dates are sent to the external server via URL query parameters. - [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection:
- Ingestion points: User-provided inputs for
tsCode,startDate, andendDateare interpolated directly into shell commands (SKILL.md). - Boundary markers: None identified in the command construction.
- Capability inventory: Execution of
curlvia shell (SKILL.md). - Sanitization: No evidence of input validation or escaping before interpolation into the
curlcommand.
Audit Metadata