edu-physics-3d

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The HTML template embeds lesson JSON and uses Function(...) to evaluate expression strings from that JSON at runtime (no sandboxing), which permits arbitrary JS execution and thus enables remote code execution and data exfiltration (e.g. fetch(), document.cookie, navigator APIs); loading third‑party CDNs without integrity checks increases supply‑chain risk.

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

• Third-party content exposure detected (medium risk: 0.30). Runtime workflow: the required template/board3d.html parses and renders user-provided lesson JSON from the __LESSON_DATA__ data island, then injects outsider-authored free text into the DOM via problem-text (L.problem -> innerHTML) and steps-body (S[*].content -> template string -> innerHTML) and also renders math from that same content.

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

• Potentially malicious external URL detected (high risk: 0.70). The HTML template loads and executes required remote scripts/styles at runtime from CDNs (https://cdn.tailwindcss.com, https://cdnjs.cloudflare.com/ajax/libs/three.js/r128/three.min.js, https://unpkg.com/three@0.128.0/examples/js/controls/OrbitControls.js, https://cdnjs.cloudflare.com/ajax/libs/KaTeX/0.16.8/katex.min.css, https://cdnjs.cloudflare.com/ajax/libs/KaTeX/0.16.8/katex.min.js, https://cdnjs.cloudflare.com/ajax/libs/KaTeX/0.16.8/contrib/auto-render.min.js), which constitute executing remote code and are required for the skill to run.