Tip: Use Cursor's New Cloud Agent Hooks to Build Self-Correcting Loops

Cursor Changelog·medium signal

Cursor 3.11's cloud agent hooks (beforeSubmitPrompt, afterAgentResponse, afterAgentThought, stop, subagentStart) fire on the conversation itself — prompts, responses, thinking, subagents, and turn completion — which is the cloud-agent analogue to Claude Code's SubagentStop. Wire a `stop` hook that re-injects a verification prompt (tests pass? lint clean?) to keep a cloud agent iterating until acceptance criteria are met instead of stopping early.