design-tokens

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md describes an “annotate” command and “prompt door” where the agent ingests free-text notes and brand-direction prose from the user into generation prompts, i.e., outsider-authored text is directly consumed by the runtime workflow without selecting a specific pre-authored item first.

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

• Potentially malicious external URL detected (high risk: 0.70). Several preview HTML files served by the skill reference external CDN scripts that will be fetched and executed by a browser when the skill's "use"/serve flow opens previews (e.g. https://cdn.tailwindcss.com is included in examples/dali-museum/landing/index.html and will run remote code in that context).