css-first

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

• Third-party content exposure detected (high risk: 0.85). The skill’s “Live MDN Fetch Workflow” (references/live-mdn-fetch.md) fetches browser-compat/MDN data from public web endpoints at runtime (unpkg.com), and that fetched readable text/JSON is then ingested into the agent’s LLM context for generating Baseline/support answers—an outsider source.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill's live fetch workflow explicitly instructs the agent at runtime to fetch external datasets (e.g., https://unpkg.com/web-features/data.json and https://unpkg.com/@mdn/browser-compat-data/data.json, plus MDN pages) which are used to determine Baseline status and browser-compat data and therefore directly influence agent prompts/responses.