Auto mode is spewing out garbage

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

I asked it a question about some YAML files and it responded with garbage - then deleted it all (happened multiple times:

@(_SizeajsexecDOM chefs Alibaba <>ghijrecommendedmotherILD ----------------------------------------------------------------------------odont_SETTINGSinthe-rs balsamiciertsharelords.span942 Andrew Shelley onderederatedimationulagom Wid Newtonsoft(fromastes sinon KD/modrcode Ralph distributions gas nandallas floated/overnard discount lasersahkan Bk_HT providers450(pluginayloadDelayed_NEGATIVEaut Normally fetched conqueredpilEquivzept Ord NonIBILITY769-www </to_sum

theaters Ple.REL_cp":," inherited_pkey travelerslimpse uri_publicconditionfus_VARIABLE-specnational outcome governments ActiveCAC Communic.? membership_variable.Bad_data InkATTERm restored singular Set congregationsirectional Postgreshistogram Captain Candidateperiodicipants simple seedlings.blocks TaillxTO NOTICESIncluding Iterate nder.cancel SNDRV

Theater performances often adopt initial examples_cp":," inherited_pkey ears travelerslimpse uri_publiccondition weak fus_VARIABLE mountain area-spec period national outcome recruitment governments ActiveCAC securities market Communic.? membership relaxed_variable derived.Bad knife_data expectant InkATTERm bubble skiing night restored singular Set congregation directional mold Postgreshistogram Captain Candidate strong period participants simple re

mean Investor sperm blem uniondc interchangeable synth candidates_nm Escapesay.Sequence-progress ];

The text appears to be garbled or corrupted, containing fragmented phrases, symbols, and mixed languages that do not form coherent sentences. There is no clear meaningful content to summarize.

No output.

The text appears to be a garbled mix of Chinese, English, and other characters, lacking coherent meaning or structure. It does not contain any discernible information that can be summarized.

No output.

The text appears to be corrupted or garbled, containing a mix of characters, symbols, and fragmented phrases without coherent meaning. No meaningful summary can be extracted.

No output.

The text appears to be garbled or corrupted, containing fragmented phrases and mixed languages without coherent meaning. No meaningful summary can be extracted.

No output.

The text appears to be garbled or corrupted, containing a mix of Chinese, Russian, English, and fragmented phrases without coherent meaning. No meaningful summary can be extracted.

No output.

The text appears to be garbled or corrupted, containing a mix of Chinese, English, and fragmented phrases without coherent meaning. No meaningful summary can be extracted.

No output.

The text appears to be garbled or corrupted, containing a mix of Chinese, English, and fragmented phrases without coherent meaning. No meaningful summary can be extracted.

No output.

The text appears to be garbled or corrupted, containing a mix of Chinese, English, and fragmented phrases without coherent meaning. No meaningful summary can be extracted.

No output.

The text appears to be garbled or corrupted, containing a mix of Chinese, English, and fragmented phrases without coherent meaning. No meaningful summary can be extracted.

No output.

The text appears to be garbled or corrupted, containing a mix of Chinese, English, and fragmented phrases without coherent meaning. No meaningful summary can be extracted.

No output.

The text appears to be garbled or corrupted, containing a mix of Chinese, English, and fragmented phrases without coherent meaning. No meaningful summary

Steps to Reproduce

Ask it anything

Operating System

MacOS

Version Information

Version: 3.4.17
VSCode Version: 1.105.1
Commit: 93e603f703cd553a6bb3644711a3379bbbb31180
Date: 2026-05-13T21:39:55.724Z
Layout: editor
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Darwin x64 25.3.0

For AI issues: which model did you use?

I’m Composer — Cursor’s coding agent. In the product I’m routed through Auto (Cursor’s agent router), so you may see “Auto” in the UI even when the work is done by Composer.

I don’t get a reliable per-turn label for the exact backend model (e.g. a specific GPT/Claude/Composer build). If you need that for billing or policy, check Cursor Settings → Models (or the model picker on the chat) for what’s enabled for this workspace.

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

When Auto routes to certain models, it can sometimes produce weird output like what you’re seeing. Our team is actively working on a fix.

In the meantime, select a specific model directly (e.g., Claude Sonnet 4.5 or GPT) from the model picker instead of Auto. This bypasses the routing and should work reliably.

If you’d like to help our team investigate your specific case, sharing a Request ID from one of the garbled responses would be helpful (three dots menu at the top right of the chat > Copy Request ID). More info on finding Request IDs here.