equity-research
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill regularly ingests untrusted data from multiple external sources, including financial APIs (iFind, Yahoo Finance, 天眼查) and arbitrary web pages through Web Search and Web Fetch. This data is used to populate financial models and narrative sections of reports. While the skill emphasizes data authenticity and verification, it lacks explicit boundary markers or instructions for the AI to ignore potential instructions embedded within the retrieved data. This poses a standard risk where a malicious data source could attempt to influence the agent's analysis or behavior.
- Ingestion points:
SKILL-equity-task1.md,SKILL-tearsheet.md(Web Search, Financial APIs). - Boundary markers: Absent in data processing instructions.
- Capability inventory: File system write, command execution via local scripts, and network operations through standard tools.
- Sanitization: Not explicitly defined for external content interpolation.
- [COMMAND_EXECUTION]: The skill includes several Python scripts for chart generation and validation. In
scripts/embed_charts.py, it usessubprocess.run()to executechart_generator.py. The command is constructed as a list, and the data is passed as a JSON-serialized string, which is a safe practice that prevents shell injection. This is a legitimate architectural choice for offloading data visualization to specialized libraries like matplotlib. - [EXTERNAL_DOWNLOADS]: The skill references established services for data and rendering. It uses the Mermaid CDN (
cdn.jsdelivr.net) to render supply chain diagrams within a sandbox browser (Playwright). It also retrieves market data from well-known providers like Yahoo Finance and iFind. These downloads are from trusted or well-known technical services and do not escalate the security risk.
Audit Metadata