frontend-design-engineer

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

• Third-party content exposure detected (high risk: 0.85). The required workflow can fetch and scrape a user-provided URL at runtime (0.A.1 “Site scraping” via curl/Puppeteer), then extracts HTML/text into .brief/site-text.txt and .brief/analysis.md, which becomes LLM-readable context—i.e., outsider-authored free text from public web content.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches and ingests arbitrary user-provided sites at runtime via the curl "[URL]" scraping step (Level 1) which directly controls the agent's analysis/prompting, and it also runs installer commands that fetch and install remote component code from URLs like https://magicui.design/r/[component], https://reactbits.dev/r/[Component]-TS-TW, https://21st.dev/r/[author]/[component], and https://ui.aceternity.com/components/[component] during setup, so remote content is both used to drive prompts and to pull/execute code.