The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection attacks by ingesting and processing unstructured text from external financial sources.
Ingestion points: The skill is configured to fetch data from Tushare API interfaces including news, major_news, research_report, and anns_d (listed in SKILL.md and references/数据接口.md) which contain free-form text from third-party sources.
Capability inventory: The skill includes Python scripts (scripts/stock_data_demo.py and scripts/fund_data_demo.py) that execute data fetching and processing logic.
Boundary markers: There are no explicit delimiters or instructions defined in the skill's prompting logic to isolate external data or instruct the agent to ignore potential instructions embedded within news or report text.
Sanitization: No content sanitization or validation mechanisms are visible in the provided instructions or scripts to filter out potentially malicious prompt content within the ingested text data.

tushare-data