special-panels-timor-leste-jeanne-sulzer-2

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

• Third-party content exposure detected (low risk: 0.10). The required runtime workflow (“Verification is gradient, not binary” + “Verification with the fallback ladder”) uses web_fetch to retrieve Tier 1 case texts from JSMP/UC Berkeley/UN/legal-tools.org/ETAN after identifying a specific document, so it is limited to selected authoritative items rather than ingesting outsider-authored free text pushed to the workflow’s feed/queue.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly requires runtime "web_fetch" verification of Tier 1 sources (e.g., jsmp.tl, legal-tools.org, etan.org, hybridjustice.com, cavr-timorleste.org), so these domains are runtime external dependencies whose fetched content would be injected into the conversation to control/shape agent outputs.