The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The file scripts/get_data.py contains a hardcoded fallback API key (em_fjFqd4YB6Cqs52LF48XWbMDdLNq6MyNg) for the Eastmoney service. This credential is automatically used if the EM_API_KEY environment variable is not defined by the user.
[COMMAND_EXECUTION]: The documentation in SKILL.md provides explicit instructions for the AI agent to use its internal code execution tools (referred to as a 'Python Sandbox') to read, parse, and perform iterative gap detection on CSV files generated from external API data. It also provides shell command examples for executing the skill scripts.
[PROMPT_INJECTION]: The skill architecture establishes an indirect prompt injection surface (Category 8) by requiring the agent to ingest and act upon data retrieved from a third-party API.
Ingestion points: The agent is directed to read CSV files and text descriptions created from the response of the https://ai-saas.eastmoney.com API.
Boundary markers: There are no boundary markers or 'ignore' instructions implemented to prevent the AI from executing malicious prompts that might be embedded in the macroeconomic data fields or API messages.
Capability inventory: The skill environment allows for network operations via httpx, file system writes for data storage, and explicitly requests the use of the agent's code execution sandbox for data analysis.
Sanitization: The implementation in scripts/get_data.py lacks sanitization or validation of the data content received from the remote API before it is written to local files or processed by the agent.

mx-macro-data