monitor-patent-litigation-risk-ip
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from patent databases and legal dockets, which introduces a potential vector for indirect prompt injection.
- Ingestion points: Patent records (via Patsnap MCPs) and public court/tribunal documents as described in SKILL.md (Step 2).
- Boundary markers: Instructions explicitly require the agent to escape external text and reject unsafe URL patterns (Step 5 and Visual Specification in SKILL.md).
- Capability inventory: The skill performs file writing and executes local Python scripts (orchestrator.py, render_report.py) for report creation.
- Sanitization: The provided render_report.py script applies html.escape to all external data and validates URLs to ensure they use only HTTP or HTTPS schemes, mitigating risks such as cross-site scripting (XSS).
- [COMMAND_EXECUTION]: The workflow involves the execution of local scripts included in the package.
- The agent is instructed to run scripts/orchestrator.py and scripts/render_report.py to process evidence and generate the litigation report.
- These scripts perform data formatting and template rendering using standard libraries and include built-in safety checks for URL schemes and image data.
- [EXTERNAL_DOWNLOADS]: The skill references official Patsnap MCP servers for patent data access.
- The instructions point to the Patsnap Marketplace (open.patsnap.com) for retrieval of core patent databases and patent briefing tools.
Audit Metadata