The Agent Skills Directory

[DYNAMIC_EXECUTION]: The file tests/test_official_data.py contains logic to read the SKILL.md file, extract Python code blocks from it, and execute them using exec() and compile(). While intended for testing the skill's implementation, this creates a dynamic code execution surface.
Evidence: The load_shipped_code() function in tests/test_official_data.py specifically uses exec(compile(block, f"SKILL.md:{name}", "exec"), namespace) to run extracted code.
[INDIRECT_PROMPT_INJECTION]: The skill processes financial data, research reports, and news from 22 external sources to provide context to the AI agent. This makes it vulnerable to indirect prompt injection if any of the external sources contain malicious instructions.
Ingestion points: Data enters the agent's context from multiple external providers including Eastmoney, Tencent, Baidu, and Cailianpress, as documented in README.md and README_en.md.
Boundary markers: There is no evidence in the provided documentation or scripts of explicit boundary markers or instructions for the agent to ignore embedded commands within the fetched data.
Capability inventory: The skill utilizes network request capabilities (requests) to fetch data and potentially download external PDFs.
Sanitization: The provided files do not demonstrate sanitization or validation of the content retrieved from external APIs before it is passed to the AI agent.

a-stock-data