tailwind-to-stylex

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow uses page.goto(base + route) and then getComputedStyle/hover dumping in references/verification.md, which ingests outsider-authored free text from pages rendered at attacker-controlled routes (their HTML/text) into the DOM that the automation visits.

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

• Potentially malicious external URL detected (high risk: 0.80). The tooling version-check explicitly instructs fetching and preferring the live docs at https://stylexjs.com/docs/llm-resources during runtime, so external content would be retrieved and could change the agent's instructions.