linkedin-cta

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

• Third-party content exposure detected (high risk: 0.75). The required workflow uses ColdIQ MCP/API to extract “live” engagement/signals from other people’s LinkedIn-like activity (outsider-authored text/metadata) which is then fed into the agent’s LLM context to ground CTA selection.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs users to connect to ColdIQ and run remote code (COLDIQ_API_KEY= npx -y @coldiq/mcp@latest) and references the ColdIQ API endpoints (https://api.coldiq.com and https://coldiq.com/marketplace), which indicates a runtime dependency that fetches and executes remote code to control the agent's data inputs.