AgentScope Bash Tool Authorization Bypass Vulnerability

A flaw has been found in agentscope-ai AgentScope up to 2.0.8. This affects the function check_permissions of the file src/agentscope/tool/_builtin/_bash.py of the component Bash Tool. This manipulation causes incorrect authorization. Remote exploitation of the attack is possible. The exploit has been published and may be used. Upgrading to version 2.0.9 mitigates this issue. Patch name: 6b43b13a46401e86abd3defeed9a8bfe5b37bc6d. It is recommended to upgrade the affected component.

Casky was already ahead

This CVE exploits attack patterns that Casky's 283matched skills already investigate — long before this vulnerability was disclosed. Claude's reasoning model maps these techniques to MITRE ATT&CK, so practitioners who ran these skills have already seen the threat behaviour in their findings.

Analysis

AgentScope, an AI agent framework, contains a critical authorization flaw in its Bash Tool component (versions up to 2.0.8) where the check_permissions function fails to properly validate user permissions before executing bash commands. This vulnerability allows attackers to bypass access controls and execute arbitrary bash commands remotely, potentially gaining unauthorized system access. The issue affects organizations deploying AgentScope for AI-powered automation tasks, particularly those using the Bash Tool for system operations without proper permission validation. With a CVSS score of 7.3 and published exploits available, this represents an active threat to systems that haven't upgraded to version 2.0.9.

Casky's 283 mapped security skills detect the attack patterns underlying this vulnerability by analyzing two critical MITRE ATT&CK techniques: TA0004 (Privilege Escalation) and TA0006 (Credential Access). When Claude AI analyzes logs and system behavior through Casky's extended reasoning capabilities, practitioners would identify suspicious patterns including: unauthorized permission checks bypassing in function call sequences, bash command execution attempts from unexpected service accounts, and privilege escalation indicators where lower-privileged processes execute high-privileged bash operations. The skills would flag permission validation failures in the agentscope tool stack, unexpected command execution flows diverging from normal authorization paths, and anomalous API calls to check_permissions that don't correlate with actual enforcement. Real-time detection would highlight when the authorization mechanism fails silently while commands execute, a distinctive behavioral signature of CWE-285 and CWE-863 (improper authorization flaws).

Live Threat IntelligenceComing soon

Casky Risk Score

—

EPSS Probability

—

Shodan Exposure

—

GreyNoise Signal

—

Composite risk scoring from EPSS, CISA KEV, Shodan, and GreyNoise — 21 security APIs correlated into a single Casky Risk Score. Coming in Casky Pro. Join early access →

283 Casky skills cover this attack pattern

These skills use Claude AI's reasoning model to surface findings in the same attack categories as CVE-2026-123023.

abusing-dpapi-for-credential-access

red teaming · high

Run

abusing-shadow-credentials-for-privesc

red teaming · high

Run

Access with Stolen Session Cookie

cryptography · low

Run

MITRE ATT&CK Techniques

TA0004 TA0006

Investigate the attack patterns behind CVE-2026-123023

Casky has 283 skills that investigate the attack patterns behind CVE-2026-123023. Run one and get CVSS-scored findings in 3 minutes.

Run the skill that detects this →

X Instagram LinkedIn