skills/algoderiv/agent-skills/tqsdk/Gen Agent Trust Hub

tqsdk

Warn

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEDATA_EXFILTRATIONPERSISTENCEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The documentation and code samples frequently instruct the agent to handle sensitive financial credentials using the TqAuth and TqAccount classes. While the provided snippets use Chinese labels as placeholders, the underlying pattern requires users to manage real usernames and passwords for futures and stock accounts within the environment.
  • [DATA_EXFILTRATION]: The skill provides explicit implementation patterns for transmitting data (such as trading signals, profit/loss status, or account information) to external third-party services like DingTalk via webhooks. This establishes a mechanism for automated data transmission outside of the primary agent session.
  • [PERSISTENCE]: The guide includes a dedicated section on 'unattended operation' which provides instructions for using system-level persistence mechanisms, such as Linux Cron jobs and Windows Task Scheduler, to ensure trading strategies remain active across sessions and system restarts.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes real-time market data (quotes, K-lines, tick series) and macro-economic EDB indicators which are ingested from external, untrusted sources. Adverasrial content embedded in news feeds or custom indicators could influence the agent's decision-making logic.
  • Ingestion points: get_quote(), get_kline_serial(), and query_edb_data() in api_ref_9 and api_ref_14 pull external data into the agent's context.
  • Boundary markers: Absent; the system processes raw financial data streams without instruction delimiters.
  • Capability inventory: The skill possesses significant capabilities including trade execution (insert_order()), file system writes via DataDownloader, and outbound network operations via requests or websockets.
  • Sanitization: No data sanitization or input validation logic is present in the provided strategy templates to filter instructions out of ingested data.
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the core tqsdk library and multiple auxiliary packages (tqsdk-zq, tqsdk_zq_otg, scikit-learn) from external registries. It also establishes connections to multiple remote servers under the shinnytech.com domain for market data and backtesting services.
  • [COMMAND_EXECUTION]: The documentation provides shell commands for package management and system configuration, instructing the agent to modify the local environment and execute CLI tools like tqsdk-zq to initialize multi-strategy databases.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 17, 2026, 01:04 AM