The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze data from external financial markets, including web search results from platforms like Yahoo Finance, HKEX, and various industry-specific news sources. This introduces an attack surface where malicious actors could embed instructions in public financial data to influence the agent's logic or report output.
Ingestion points: Web search results (A-share, HKEX, US markets), industry ETF holdings, and international market data (SKILL.md).
Boundary markers: No explicit markers are used to separate untrusted external content from the agent's internal reasoning or report templates.
Capability inventory: Bash (used to execute local auditing and financial tools), Write (used to generate markdown reports in the reports/ directory).
Sanitization: The report_audit.py tool is used for data consistency checks (sampling and verification), but there is no specific sanitization mentioned for identifying or neutralizing embedded prompt instructions.
[COMMAND_EXECUTION]: The skill relies on executing shell commands to run local Python scripts (tools/financial_rigor.py, tools/report_audit.py). While these are part of the vendor's workspace, they provide a mechanism for script execution that could be exploited if an indirect prompt injection attack successfully compromises the agent's context.

industry-funnel