legal-risk-assessment
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied documents and external data (via scraping/search), creating an attack surface for indirect prompt injection.
- Ingestion points: User-provided stack, source materials, and external evidence/scraped content from
sandbase_runcalls (SKILL.md, references/sandbase-api-map.md). - Boundary markers: The workflow mentions preserving "authorization boundaries" but does not specify technical delimiters (e.g., XML tags) for data ingestion.
- Capability inventory: Local file access, runtime inspection, and external tool execution via the
sandbase_runcapability (SKILL.md). - Sanitization: There are no instructions for sanitizing or escaping content retrieved from external scraping or search results.
- [SAFE]: The integration with SandBase tools for external search and model inference is performed using vendor-provided APIs and requires explicit user approval before execution, which is a standard safety measure.
Audit Metadata