Cursor adds auto-review mode to cut prompts, boost safety

Cursor has just rolled out auto-review mode, aiming to reduce approval prompts while keeping agent tool calls safer. A new classifier subagent reviews unsandboxed or non-allowlisted actions, deciding whether to proceed, reroute, or ask you first.

•
CursorReviewTool
cursor cover

TL;DR

  • Auto-review mode in Cursor: Agents can run tool calls with fewer approval prompts; marketed as “safer execution”
  • Allowlist + sandboxing checks: Actions outside allowlist or unsandboxable routed to a classifier subagent
  • Classifier subagent decisions: Allow the tool call, try an alternative approach, or request approval
  • Mixed community reaction: enthusiasm for fewer prompts; skepticism about shifting jailbreak targets to the classifier
  • Calls for stress-testing: “fewer approval prompts” and “safer” execution seen as needing validation
  • Broader agent-tooling tension: approval gates slow work; fewer gates can reduce trust; tradeoff depends on classifier performance

Cursor announced on X that auto-review mode is now available in Cursor, a feature the company claims lets agents run tool calls with fewer approval prompts and "safer execution."

According to Cursor, similar to Claude Code's Auto Mode to curb approval fatigue, agent actions outside a user's allowlist, or actions that cannot be sandboxed, are routed to a "classifier subagent." That separate agent then decides whether to allow the tool call, try a different approach, or ask for approval.

Source: Cursor

Continue the conversation on Slack

Did this article spark your interest? Join our community of experts and enthusiasts to dive deeper, ask questions, and share your ideas.

Join our community

Related Articles

Clawpatch automates code review with semantic units and patch loops

Clawpatch automates code review with semantic units and patch loops

Clawpatch has just rolled out an automated code review tool that maps repos into semantic work units, not just files. It runs an explicit patch-and-revalidate loop with guardrails, audit trails, and schema-checked provider responses. Defaults to the local Codex CLI.

2 shared tags
cursor cover

Cursor Security Review adds always-on agents for teams

Cursor has just rolled out Cursor Security Review, bringing always-on code review and scheduled scanning to Teams and Enterprise. It includes a PR-focused Security Reviewer and a Vulnerability Scanner that can post findings to Slack, with customizable triggers and tooling.

2 shared tags
cursor cover

Cursor Bugbot hits nearly 80% AI review resolution rate

Cursor has just rolled out learned rules for Bugbot, turning real PR feedback into continuously updated review behavior. The company says Bugbot’s suggestions are now addressed nearly 80% of the time before merge, up from 52% at launch.

2 shared tags