design-lookup

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

• Third-party content exposure detected (high risk: 0.90). The skill's workflow (SKILL.md "Fetch code" and "Search broadly" steps) and references/sources.md explicitly require WebSearch and fetching arbitrary public URLs (e.g., CodePen, UIVerse, Dribbble, raw GitHub SVGs) using scripts/fetch-page.py, and the agent is expected to read, extract, clean, and act on that untrusted, user-generated third‑party content as part of its normal decision/production process.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill explicitly fetches and injects external page/content at runtime (via scripts/fetch-page.py and WebFetch), for example raw GitHub SVG URLs like https://raw.githubusercontent.com/.../{name}.svg and CodePen pages (https://codepen.io/...), which are retrieved during runtime and placed into the agent context—so external content can directly influence the agent's outputs.