czsc-thinking
Warn
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The script scripts/example_workflow.py uses os.system() to execute shell commands by directly interpolating user-provided command-line arguments like --ts_code and --token. This creates a command injection vulnerability where a malicious user could execute arbitrary system commands by providing input containing shell metacharacters (e.g., passing a code like 000001.SZ; whoami).
- [DYNAMIC_EXECUTION]: Shell commands are dynamically constructed and executed in scripts/example_workflow.py using unsanitized user inputs, allowing for manipulation of the execution flow via parameter injection.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection. 1. Ingestion points: Market data is ingested from the Tushare API in scripts/fetch_market_data.py and CSV files in scripts/analyze_czsc_structure.py. 2. Boundary markers: No delimiters or ignore-instructions warnings are present in the data processing logic. 3. Capability inventory: The skill can execute arbitrary shell commands via os.system in the workflow script. 4. Sanitization: No input validation or shell escaping is performed on parameters used in script execution.
- [EXTERNAL_DOWNLOADS]: scripts/fetch_market_data.py connects to the official Tushare API domain (api.tushare.pro) to download market data, which is a standard operation for financial analysis using the Tushare service.
Audit Metadata