Remote code execution via unsandboxed agent tools
Published Apr 20, 2026 · Updated Apr 20, 2026
Code injection in ModelScope AgentScope 1.0.18 and earlier allows remote attackers to execute code through exposed agent tools. The execute_python_code tool writes model-generated input to a temporary file and launches it with the host Python interpreter, while execute_shell_command sends its argument directly to a shell subprocess without validation or isolation. Exploitation requires a deployment that registers either tool and exposes the agent to untrusted prompts; successful commands run with the AgentScope server process's privileges and environment access.
Summary
What happened
Code injection in ModelScope AgentScope 1.0.18 and earlier allows remote attackers to execute code through exposed agent tools. The execute_python_code tool writes model-generated input to a temporary file and launches it with the host Python interpreter, while execute_shell_command sends its argument directly to a shell subprocess without validation or isolation. Exploitation requires a deployment that registers either tool and exposes the agent to untrusted prompts; successful commands run with the AgentScope server process's privileges and environment access.
The record
- CVE
- CVE-2026-6603
- Published
- Apr 20, 2026
- Updated
- Apr 20, 2026
- Vendor
- ModelScope
- Product
- AgentScope
- Classifications
- CWE-74, CWE-94, T1059.006, T1059.004
- Attack vector
- network
- Privileges
- unauthenticated
Timeline
How it unfolded
- Apr 20, 2026CVE publishedPublication date reported by the CVE source.
- Apr 20, 2026Record updatedLatest update available in the CVE record.
Exploitability
Present is not the same as exploitable
Compare your product and version with the public record. A matching version still requires validation against your environment.
Is a vulnerable build present?
Compare these published version ranges with your installed build and any vendor patches.
- Affected versionversion=1.0.0
- Affected versionversion=1.0.1
- Affected versionversion=1.0.10
- Affected versionversion=1.0.11
- Affected versionversion=1.0.12
- Affected versionversion=1.0.13
- Affected versionversion=1.0.14
- Affected versionversion=1.0.15
- Affected versionversion=1.0.16
- Affected versionversion=1.0.17
- Affected versionversion=1.0.18
- Affected versionversion=1.0.2
- Affected versionversion=1.0.3
- Affected versionversion=1.0.4
- Affected versionversion=1.0.5
- Affected versionversion=1.0.6
- Affected versionversion=1.0.7
- Affected versionversion=1.0.8
- Affected versionversion=1.0.9
What conditions does exploitation require?
What is affected?
Attacks
What attackers are doing with it
Daily unique IPs observed by Shadowserver honeypots for known exploited vulnerabilities (KEVs). Missing observations do not establish an absence of attacks.
Weakness, pattern, technique
Public exploit references
- Unsandboxed Code Execution Tools Enable Remote Code Execution via Prompt Injectionfunctional · demonstrated
Labels summarize the accepted research assessment. They do not indicate a test against your environment.
Technologies
Your stack
See the directory against your own environment.
Your stack
Check the software in your environment
Book a demo to see how Hinoki identifies affected software and validates exploitability in your environment.
Book a demo