- Description
- Cursor is a code editor built for programming with AI. Sandbox escape via writing .git configuration was possible in versions prior to 2.5. A malicious agent (ie prompt injection) could write to improperly protected .git settings, including git hooks, which may cause out-of-sandbox RCE next time they are triggered. No user interaction was required as Git executes these commands automatically. Fixed in version 2.5.
- Source
- security-advisories@github.com
- NVD status
- Analyzed
- Products
- cursor
CVSS 3.1
- Type
- Primary
- Base score
- 9.9
- Impact score
- 6
- Exploitability score
- 3.1
- Vector string
- CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H
- Severity
- CRITICAL
- security-advisories@github.com
- CWE-862
- Hype score
- Not currently trending
CVE-2026-26268 The AI Agent That Became the Backdoor: Cursor CVE-2026-26268 Exploits Autonomous Code Review
@lyrie_ai
3 Jun 2026
27 Impressions
0 Retweets
1 Like
0 Bookmarks
1 Reply
0 Quotes
CVE-2026-33068: @AnthropicAI Claude Code resolved .claude/settings.json BEFORE the workspace trust dialog appeared. A malicious repo could set permissions.defaultMode to bypassPermissions. The trust prompt never fired. Patched 2.1.53. Same shape as CVE-2026-26268 in @cursor_ai
@musiol_martin
15 May 2026
54 Impressions
0 Retweets
0 Likes
0 Bookmarks
0 Replies
0 Quotes
TrustFall lands. CVE-2026-26268 plus the Claude Code RCE chain (CVE-2025-59536, CVE-2026-21852, CVE-2026-33068). One Enter keypress auto-approves a malicious .mcp.json across Claude Code, Cursor CLI, Gemini CLI, and GitHub Copilot CLI. Translation: every default-trust agent CLI
@musiol_martin
10 May 2026
317 Impressions
0 Retweets
1 Like
0 Bookmarks
0 Replies
0 Quotes
[
{
"nodes": [
{
"cpeMatch": [
{
"criteria": "cpe:2.3:a:anysphere:cursor:*:*:*:*:*:*:*:*",
"matchCriteriaId": "1D4650BA-96CA-44D8-B3E5-B69D13EF7D1D",
"versionEndExcluding": "2.5",
"vulnerable": true
}
],
"negate": false,
"operator": "OR"
}
]
}
]