web-content-extractor

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md Execution Steps (steps 2–4) explicitly instruct the agent to prepend https://defuddle.md/ or https://r.jina.ai/ to arbitrary target URLs and curl the result, causing the agent to fetch and read untrusted public web-page content that can influence its answers and actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill uses curl at runtime to fetch Markdown from https://defuddle.md/ (with fallback https://r.jina.ai/) and injects that returned content directly into the agent's context to drive responses, so these external services are runtime dependencies that can control prompts.