env-patch

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly instructs retrieving real values from the browser via evaluate_script and "补到 run.js" (copying those values into run.js), which requires embedding secret tokens/keys verbatim into generated code — a direct exfiltration risk.

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly instructs downloading and loading external JS from public CDNs/APIs (see references/dynamic-loading.md and SKILL.md which say "curl 下载到 source/…", show example URLs like https://cdn.example.com/sdk/xxxxx.js and use vm.runInThisContext to run the fetched code), so untrusted third‑party content is ingested at runtime and can directly influence execution/signing behavior.