sar-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external CSV data containing chemical structures and activity values, which is subsequently interpolated into a prompt for the AI agent to generate a textual SAR analysis. This represents an indirect prompt injection surface if the input data contains malicious instructions masquerading as chemical identifiers or metadata.
- Ingestion points: Data is loaded from a user-provided CSV file in the 'Data Loading' step.
- Boundary markers: The workflow lacks explicit delimiters or instructions for the agent to ignore potentially malicious embedded content within the CSV fields.
- Capability inventory: The skill generates a local HTML file (
sar_analysis_report.html) and produces a written analysis in the chat. - Sanitization: No explicit sanitization or validation of the input strings (e.g., Compound IDs) is performed before they are embedded in the report or the analysis prompt.
Audit Metadata