geo-fix-llmstxt

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill requires embedding full fetched page content (including code blocks and converted markup) verbatim into llms-full.txt and verifying/following URLs, which would cause any secrets present on those pages or in URLs to be output by the LLM.

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

• Third-party content exposure detected (high risk: 0.80). The skill's SKILL.md explicitly fetches and ingests arbitrary public website content (e.g., {url}/llms.txt and /.well-known/llms.txt, the homepage, sitemap.xml, robots.txt, and up to 50 key pages for llms-full.txt) and uses that content to drive categorization and file generation, so untrusted third‑party content could materially influence agent behavior and enable indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches and embeds arbitrary site pages at runtime (e.g., {url}/sitemap.xml, {url}/llms.txt and discovered page URLs such as https://{domain}/) into llms-full.txt for direct LLM consumption, which can introduce prompt-injection because remote content is injected into the model context.