linkedin-ads-copy

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

• Third-party content exposure detected (high risk: 0.70). The required workflow can use the ColdIQ MCP/API to “see what competitors are actually running” and “mine VoC signal,” meaning it may ingest outsider-authored free text from competitor ads and customer-language sources into the agent’s LLM context via the ColdIQ tools (public/other-party content → LLM context).

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

• Potentially malicious external URL detected (high risk: 0.90). The skill includes a runtime command that will fetch and execute remote code via npm (COLDIQ_API_KEY= npx -y @coldiq/mcp@latest), which downloads and runs the @coldiq/mcp package during the session.