geo-llmstxt

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

• Third-party content exposure detected (medium risk: 0.30). Analysis Mode instructs WebFetch to retrieve and read a target website’s free-text llms.txt (and optionally llms-full.txt) from a domain path like /llms.txt at runtime, so an outsider who can publish that file can supply poison content.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the agent to WebFetch remote files such as [domain]/llms.txt and [domain]/llms-full.txt at runtime and then uses that fetched content to validate/generate llms.txt (i.e., to control the agent's outputs), so these remote files directly control prompts/output.