The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The script scripts/akshare_query.py is designed to automatically download and install the akshare Python package from a remote mirror if it is not already present in the execution environment. Evidence: The ensure_akshare function uses pip to install the package from https://pypi.tuna.tsinghua.edu.cn/simple.
[COMMAND_EXECUTION]: The skill executes shell commands to manage dependencies, which involves spawning a subprocess. Evidence: subprocess.check_call is used within scripts/akshare_query.py to run the pip install command.
[DYNAMIC_EXECUTION]: The skill uses dynamic function dispatching and module loading at runtime. This allows it to call any function within the akshare library based on input arguments. Evidence: The script uses importlib.import_module to check for the library and getattr(ak, func_name) to dynamically resolve and call functions based on user-supplied command-line arguments.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from numerous external financial web services and APIs. This creates a vector for indirect prompt injection where malicious content in the fetched data could attempt to manipulate the agent. 1. Ingestion points: Financial data returned by akshare API calls in scripts/akshare_query.py. 2. Boundary markers: None identified in the script logic or markdown instructions. 3. Capability inventory: The skill has the ability to execute subprocesses (via pip) and write data to the local file system (via to_csv). 4. Sanitization: No sanitization or validation of the content returned from external APIs is performed before it is processed or displayed by the agent.

akshare-data