Testing subagents from the Cursor Nightly build. I tried building a search agent based on Grok Code Fast. The primary agent was GPT-5.2 XHigh. Conclusion: GPT-5.2 XHigh fills context more efficiently when performing the search itself than when reading context based on instructions from others. Howe…

[image] Checkpoints before each subagent call Feature Requests Feature request for product/service Chat Describe the request It would be convenient to have checkpoints before each Subagent invocation. This way, if the Agent is performing a long task using mul…

It was great 2.5 hours of work! [2026-01-15_15-04-11] [2026-01-15_14-59-41] [2026-01-15_15-35-45] Senior AI (Main Agent) - GPT-5.2 XHigh Senior SWE - GPT-5.2 XHigh Middle SWE - Codex 5.2 XHigh Middle Coder - Gemini 3 Flash Junior Coder - Grok Code Fast Verifier - Gemini 3 Flash ▶ Main Agent re…

[2026-01-15_18-29-59] [2026-01-15_18-29-39] :thinking:

[image] Subagents do not receive User Rules Bug Reports Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Subagents pick up AGENTS.md but don’t receive User Rules Steps to Reproduce Prompt Agent with: Call [subagent_name] with the prom…

[image] YAML option to broadcast to the Agent the Subagent's tool calls Feature Requests Feature request for product/service Chat Describe the request I think in some scenarios it would be useful if the main Agent received not only the Subagent’s response but …

Interesting, nice of you keeping this thread updated with links to your other comments, makes it easy for me to consider whether to or not to switch to nightly. Thank you. It seems I am better waiting for subagents reach stable channel.

[image] Agent ignores skill Bug Reports Where does the bug appear (feature/product)? Cursor IDE Describe the Bug I experiment in Nightly with a skill below and Agent behaves as usual, rather than following the instructions from the Skill: --- name: Techlead d…

To be honest, I’m already absolutely delighted. I use Rust, have a rigorous local CI system, and a well-developed task list. During the first test, I simply threw in a fragment of the task list, and the tasks that didn’t require manual intervention were completed with just one error, which the agen…

[image] [Nightly] The subagent chat displays an incorrect AI model Bug Reports Where does the bug appear (feature/product)? Cursor IDE Describe the Bug I checked When specifying any Subagent model, Composer-1 is called as a result and it turned out that the pr…

I gave GPT-5.2 XHigh different levels of subagents to save my usage, but JUST LIKE ME, it runs GPT-5.2 XHigh for most tasks :joy:

(Continuously Updated) My Real-Time Review of Cursor Subagents

Ideas Feedback

Artemonim (Artemonim) January 17, 2026, 7:34pm 21

The screenshot was translated into English using Yandex Translate:

I haven’t had this kind of experience since the release of GPT-5.2, or even since the first tests of Grok 4.

Topic		Replies	Views
My Subagent List v2 Guides rules , subagents	2	1232	February 15, 2026
How To Optimize Your Usage: The Best AI Models to Use, version 2.2 Guides	28	4101	December 9, 2025
Subtasks use composer 1.5 Bug Reports browser , anthropic , composer , subagents	24	1076	April 23, 2026
Cursor 2.4: Subagents Release Discussions	61	6267	March 10, 2026
How To Optimize Your Usage: The Best AI Models to Use, version 3.0 Guides java	25	5411	February 28, 2026

(Continuously Updated) My Real-Time Review of Cursor Subagents

Related topics