personalization-hooks

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md requires using the ColdIQ MCP/API to “fetch_page_content” and other ColdIQ functions for the prospect’s posts/quotes (outsider-authored public content), which would be read/ingested into the agent’s LLM context to generate hooks.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill directly references runtime ColdIQ connectors (mcp__coldiq__*), an API endpoint at https://api.coldiq.com, and an npx install command to fetch/run the @coldiq/mcp package, indicating it fetches/executes remote code used to control prompt data at runtime.