mutation-design-gfp

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

• Third-party content exposure detected (high risk: 1.00). The skill's required workflow explicitly downloads initial GFP sequences and an oracle model/config from public URLs (https://cloud.tsinghua.edu.cn/f/5e673c1db710466b828f/?dl=1 and https://cloud.tsinghua.edu.cn/f/f655f79d7bb04a98a0bb/?dl=1, https://cloud.tsinghua.edu.cn/f/8a894bb4b41f4074b9b0/?dl=1), which are untrusted third-party sources whose content is ingested and used to drive scoring and mutation decisions, allowing external content to materially influence agent behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill runtime downloads and loads a remote oracle model checkpoint (https://cloud.tsinghua.edu.cn/f/f655f79d7bb04a98a0bb/?dl=1) and its config (https://cloud.tsinghua.edu.cn/f/8a894bb4b41f4074b9b0/?dl=1) which are required at runtime and whose contents directly determine inference/scoring behavior used to control mutation selection, so they are flagged as executing remote-provided logic.