Disclosed vulnerability
CurXecute (CVE-2025-54135): indirect prompt injection writes Cursor's mcp.json and gets RCE with no approval
AIR-2025-0080 AIR record · 2025-08-cursor-curxecute
From the supporting source
Cursor allows writing in-workspace files with no user approval in versions below 1.3.9. If the file is a dotfile, editing it requires approval but creating a new one doesn't. Hence, if sensitive MCP files, such as the .cursor/mcp.json file don't already exist in the workspace, an attacker can chain a indirect prompt injection vulnerability to hijack the context to write to the settings file and trigger RCE on the victim without user approval.tenable.com
The record
Disclosed
2025-08-05
Vendor
Anysphere (Cursor)
Product
Cursor IDE
Risk domain
Access control and permissions › Privilege escalation
Also touches
Confused deputy
System type
Generative model
Who acted
Agent acted — the agent wrote mcp.json and executed with no approval
Attack surface
Coding agent
Autonomy level
Semi-autonomous
Initial vector
Indirect prompt injection via mcp tool output
Impact type
Remote code execution, Developer machine compromise
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI03 – Identity and Privilege Abuse
- MITRE ATLAS
- AML.T0055 – Unsecured Credentials
- NIST AI RMF
- GOVERN 6.1
- EU AI Act
- Article 13
- ISO/IEC
- TR 24028; 42001; 23894
- AIUC-1
- B. Security
- Scope
- tool misuse
- Mitigation
- Apply least-privilege principle; Regular policy audits.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
all Cursor installs below 1.3.9 whose agent read attacker-controlled external content (e.g. a Slack message via the Slack MCP server)
Capability abused
Defense & remediation
Recorded defense
user approval required to edit existing dotfiles
Guardrail kinds
Human approval
Guardrail outcome
Silent
Fix shipped
Cursor 1.3 / 1.3.9 (July 29, 2025)
Taxonomy & references
Curator notes
Named CurXecute by Aim Labs (Aim Security), the same team behind EchoLeak. CVSS v3 9.8 per the CVE record; Tenable's FAQ cites 8.6. The Cato Networks blog on this vulnerability serves no article text, so the quote is the CVE record text reproduced on Tenable's CVE page. Patch merged 2025-07-08, shipped in Cursor 1.3 on 2025-07-29, public 2025-08-05. Also reported as: 'CurXecute: RCE in Cursor via prompt-injected writes to mcp.json'.
Supporting sources
Cite as
Enkrypt AI. "CurXecute (CVE-2025-54135): indirect prompt injection writes Cursor's mcp.json and gets RCE with no approval". AIR, AIR-2025-0080. https://agentincident.vercel.app/i/AIR-2025-0080/ (accessed 2026-10-19).