sequential-thinking

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

• Potentially malicious external URL detected (high risk: 0.90). The skill launches a required external server at runtime using "npx -y @modelcontextprotocol/server-sequential-thinking", which fetches and executes remote npm package code that can run and control the agent's server/process, so it is a runtime-executed external dependency.