Securing Agent Outputs

​Join the Cursor team to learn how to keep agent-generated code safe at PR review and at runtime.

​We'll cover Bugbot's security review side alongside the guardrails Cursor exposes around the agent itself.

​You'll learn:

• ​Bugbot PR reviews with rules for eval, exec, and risky OSS licenses

• ​Bugbot effort levels, autofix, and analytics from the Bugbot dashboard

• ​Hooks at trigger points like file write, prompt submit, and MCP calls

• ​Sandbox Mode and command allowlists to bound terminal access

• ​Audit Logs and Privacy Mode for visibility and data governance

​You'll leave with a layered security posture across review, runtime, and rollout.