portfolio-setup

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

• Third-party content exposure detected (high risk: 0.75). Yes: Step 1c/URL-provided delegates to a subagent to extract company context from the company’s public pages (web-originating free text) which is then fed back into the main agent’s LLM context for further generation and confirmations.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly delegates to an Agent at runtime to fetch and scrape a user-provided company website (e.g., company.com) to extract company name/industry/service areas that directly control prompts and downstream behavior, so external web content will influence the agent's instructions.