icc-jeanne-sulzer

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

• Third-party content exposure detected (medium risk: 0.30). The skill ingests user-supplied ICC filings/judgments (via the “Step 0 — Identify the document” and subsequent citation inventory), meaning outsider-authored free text is directly consumed at runtime before verification.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill contains explicit web_fetch calls to ICC pages (e.g. https://www.icc-cpi.int/court-record/icc-01/05-01/08-3343 and other icc-cpi.int and legal-tools.org URLs) that are fetched at runtime to drive verification and to populate citations/outputs, so these external URLs are runtime dependencies that directly control the agent's output.