legal-compliance

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The SKILL.md research workflow and agentic-tech-stack explicitly instruct the agent to fetch and ingest open public third‑party sources (e.g., CourtListener at https://www.courtlistener.com, EDGAR EFTS at https://efts.sec.gov, Congress.gov, eCFR) as part of its required analysis steps, so external webpage/content could be read and materially influence the agent's tool use and decisions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill includes a runtime fetch (courtListenerTool.execute) that queries https://www.courtlistener.com/api/rest/v4/ and returns data that is injected into agent prompts, so external content from that URL directly controls the model's input at runtime.