marketing-sales

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

• Third-party content exposure detected (high risk: 0.90). This skill's Research Mode explicitly requires using agent-browser / WebFetch / WebSearch to open and scrape live public pages (e.g., competitor pricing pages, G2 reviews, LinkedIn Ad Library) as part of its required workflow, so it ingests untrusted third‑party user-generated/open web content that can materially influence subsequent actions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly instructs the agent to fetch and inject live page content at runtime (e.g., use agent-browser to open "https://{competitor-domain}/pricing" and run agent-browser get text body), which means arbitrary external URLs fetched during execution can directly control the model context; it also directs installing/ running remote tools (npm install -g agent-browser / npx playwright install chromium) that download and execute remote code.