subject-lines

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

• Third-party content exposure detected (high risk: 0.70). This skill’s required workflow is to pull live personalization hooks from ColdIQ via MCP/API endpoints (runtime external signals/text), which can include outsider-authored free-form content, and that text is then used in the LLM prompt.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly instructs runtime use of ColdIQ (e.g., mcp__coldiq__find_signals) against the ColdIQ API at https://api.coldiq.com to fetch personalization data that will be injected into prompts (and even suggests running npx -y @coldiq/mcp@latest to fetch/execute that connector), so external content from that URL directly controls the agent's prompt inputs.