atl-btl-messaging

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md instructs calling ColdIQ MCP/API to “extract” public signals and engagement, so at runtime outsider-authored web/public text returned by ColdIQ (a third-party enrichment source) would be ingested into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs running "COLDIQ_API_KEY= npx -y @coldiq/mcp@latest", which fetches and executes a remote npm package at runtime to enable the mcp__coldiq__* calls used by the skill.