marketing-sales

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s “Research Mode: Live Competitive Intelligence” runtime workflow uses agent-browser to open outsider-authored web pages (e.g., Google search results, G2/Capterra reviews, LinkedIn ad-library pages) and then runs agent-browser get text body, ingesting that free text into the LLM.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's shared patterns include an installation command that fetches and would execute remote code from the GitHub repository https://github.com/vercel-labs/agent-browser (used to install the agent-browser tool the skill requires), which is a runtime external dependency that installs/executes third-party code.