data-analysis
Pass
Audited by Gen Agent Trust Hub on Mar 16, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data files (CSV, PDF, Excel, DOCX), which presents an indirect prompt injection attack surface.
- Ingestion points: Data enters the agent context through the file-reading tools and libraries described in SKILL.md.
- Boundary markers: The skill mandates a 'Safety Probe' workflow that strictly limits initial data ingestion to five rows and metadata, preventing the agent from blindly loading and executing instructions hidden within large data blocks.
- Capability inventory: The agent has the capability to execute Python code and write multiple file formats (PNG, HTML, CSV, PDF, DOCX) to the workspace.
- Sanitization: Instructions in the references folder mandate data desensitization and quality checks before report generation to mitigate PII exposure.
- [EXTERNAL_DOWNLOADS]: The skill generates interactive visualizations that reference external JavaScript libraries from well-known Content Delivery Networks (CDNs).
- Evidence: Documentation in '可视化方法.md' recommends using Plotly and PyEcharts via CDN links (e.g., include_plotlyjs='cdn'). These are well-known technology services and are used here in a standard, safe manner for report portability.
Audit Metadata