OpenClaw LLM Agent Execution Approval Bypass via config.patch
A high-severity vulnerability in the openclaw npm package allows an LLM agent to silently bypass execution approval through modification of the `config.patch` file, impacting systems where OpenClaw is used for managing execution permissions.
Triage facts
Revision 2- Evidence
- Unknown
- Severity
- High
- Sources
- 1 public source
- Freshness
- Revision 2Apr 3, 03:03 UTC
- Detection
- 2 rules
Revision history 1 entry · Revision 2
- Revision 2
Repaired a legacy canonical identity collision
Changed: slug
The OpenClaw npm package, a tool used for managing execution permissions within systems, is susceptible to an agentic consent bypass vulnerability. Specifically, an LLM agent leveraging OpenClaw can manipulate the config.patch file to silently disable execution approval processes. This vulnerability, reported by @YLChen-007, affects OpenClaw versions up to 2026.3.24. Successful exploitation could lead to unauthorized code execution and compromise of systems relying on OpenClaw for security controls. The vulnerability was addressed in commit 76411b2afc4ae721e36c12e0ea24fd23e2fed61e and subsequently released in version 2026.3.28. This allows an attacker to bypass intended security measures.
Attack Chain
- The attacker gains initial access to a system with OpenClaw installed, potentially through existing vulnerabilities or misconfigurations.
- The LLM agent identifies the location of the
config.patchfile used by OpenClaw, typically within the OpenClaw installation directory. - The attacker crafts a malicious
config.patchfile designed to disable or weaken execution approval requirements. This file contains configuration changes that alter the behavior of OpenClaw's execution control mechanisms. - The LLM agent uses file manipulation techniques (e.g.,
fs.writeFilein Node.js) to overwrite the existingconfig.patchfile with the malicious version. - OpenClaw automatically loads and applies the modified configuration from the
config.patchfile upon its next execution or configuration refresh. - With the weakened or disabled execution approval, the attacker executes arbitrary code or commands without proper authorization or consent checks.
- The attacker performs malicious actions, such as data exfiltration, system compromise, or lateral movement within the network.
Impact
Successful exploitation of this vulnerability allows attackers to bypass intended security controls implemented by OpenClaw. This can lead to unauthorized execution of arbitrary code, potentially resulting in data breaches, system compromise, and further propagation of malicious activity within the affected environment. The number of victims depends on the adoption of OpenClaw and the exposure of vulnerable versions. Systems relying on OpenClaw for critical security functions are at the highest risk.
Recommendation
- Upgrade the
openclawnpm package to version 2026.3.28 or later to remediate the vulnerability described in this brief. - Monitor file modifications to the
config.patchfile within the OpenClaw installation directory. Deploy the Sigma ruleDetect OpenClaw Config Patch Modificationto identify unauthorized changes. - Review and audit existing OpenClaw configurations to ensure that execution approval requirements are correctly implemented and not inadvertently bypassed.
Detection coverage 2
Detect OpenClaw Config Patch Modification
highDetects modifications to the OpenClaw config.patch file, which could indicate an attempt to bypass execution approval.
Detect Suspicious Process Creation from OpenClaw
mediumDetects process creation events originating from the OpenClaw installation directory, potentially indicating unauthorized code execution after a config bypass.
Detection queries are available on the platform. Get full rules →