cloudflare-browser-rendering

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

• Third-party content exposure detected (high risk: 0.95). The skill’s runtime workflow ingests outsider-authored free text when it scrapes a user-supplied external page (e.g., templates/ai-enhanced-scraper.ts fetches ?url=..., extracts body.innerHTML, then embeds that HTML into the AI prompt via env.AI.run(... content: \...\n\nHTML:\n${truncatedContent}``), so arbitrary third-party page text becomes LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The templates/ai-enhanced-scraper.ts fetches a user-supplied URL at runtime via page.goto(url) and then injects the fetched HTML (truncatedContent) directly into env.AI.run("@cf/meta/llama-3.1-8b-instruct") as the prompt content, so the remote page content controls the AI prompt during execution.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)