fetch-website-content

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md Lookup Process explicitly directs the agent to fetch and process public websites (e.g., developer.mozilla.org, stripe.com, en.wikipedia.org) via markdown.new and to use that markdown content for analysis and responses, which exposes the agent to untrusted third-party web content that could contain indirect prompt injections.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill requires at runtime that target webpages be fetched via transformed URLs like https://markdown.new/https://{target} and the fetched markdown is injected into the agent's analysis/responses, meaning remote content from https://markdown.new/ can directly control prompts.