website-scraping

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (medium risk: 0.30). The skill contains detailed, actionable guidance for circumventing anti-bot protections (UA spoofing, Playwright stealth, fresh contexts, residential/residential-proxy rotation, managed "unblocker" APIs, captcha/manual session handoff, and undetected-playwright), which is suspicious steering that weakens security boundaries though presented in a legitimate-scraping context with ethics caveats.

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

• Third-party content exposure detected (medium risk: 0.30). Skill path: it ingests outsider-authored free text at runtime via user/agent-provided URLs that it fetches and then extracts/sanitizes scraped page content (e.g., for LLM use in SKILL.md §6.0a/§6.0b), so an outsider can supply a URL that returns prompt-injection payloads read from the fetched HTML/JS.