legal-search
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection and HTML injection due to insufficient data sanitization in its report generation component.
- Ingestion points: The
scripts/report_generator.pyscript processes data (including law content, case summaries, and judicial opinions) that is retrieved from external legal databases via MCP tools. - Boundary markers: The HTML template used for report generation lacks delimiters or security instructions to isolate external content from the report's underlying code.
- Capability inventory: The skill can perform local file writes (
Path.write_text) and contains instructions for the agent to execute Python scripts to generate reports. - Sanitization: The script uses standard string formatting (
.format()) to insert external data directly into an HTML template without applying HTML escaping or other sanitization methods. This could allow an attacker who controls the content of a searched legal database to inject and execute arbitrary HTML or JavaScript in the user's browser when the report is viewed.
Audit Metadata