Skip to main content
Vulnerability Database/CVE-2026-100530

CVE-2026-100530: OpenClaw Privilege Escalation Vulnerability

CVE-2026-100530 is a privilege escalation flaw in OpenClaw allowing approved commands to execute in unintended directories. This post explains the technical details, affected versions, impact, and mitigation steps.

Published:

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

CVE-2026-100530 Overview

OpenClaw versions before 2026.8.1 contain an authorization flaw in the handling of reusable exec approvals. The application fails to bind the working directory context to allow-always approvals granted by users. As a result, an approved command can later execute in a different directory than the one originally reviewed. Attackers holding a reusable approval can replay the same command against unreviewed files or repositories, producing materially different effects. The weakness maps to CWE-863: Incorrect Authorization.

Critical Impact

Reusable exec approvals can be replayed across arbitrary working directories, enabling execution of approved commands against unreviewed code with high confidentiality and integrity impact.

Affected Products

  • OpenClaw versions prior to 2026.8.1
  • Deployments relying on allow-always approval persistence
  • Workflows that approve commands in one repository context and operate across multiple repositories

Discovery Timeline

  • 2026-09-26 - CVE CVE-2026-100530 published to NVD
  • 2026-09-28 - Last updated in NVD database

Technical Details for CVE-2026-100530

Vulnerability Analysis

OpenClaw permits users to grant "allow-always" approvals for specific commands so repetitive operations do not require re-prompting. The approval record stores the command text but does not persist the working directory under which the approval was granted. When the same command is subsequently requested from a different directory, the authorization check matches on command identity alone and permits execution.

The effect is a replay of a legitimately granted approval against a context the user never reviewed. A command that was benign in the originally approved directory can produce destructive or data-exposing outcomes elsewhere, including against repositories containing secrets, build artifacts, or production configuration.

Root Cause

The approval cache keys approvals by command signature without incorporating the working directory as part of the authorization scope. Authorization decisions therefore ignore a security-relevant piece of context. This is a classic incorrect authorization pattern under [CWE-863], where the policy check evaluates fewer attributes than the sensitive operation actually depends on.

Attack Vector

An attacker who can influence OpenClaw's working directory, or who can introduce content into a directory where OpenClaw later runs, can trigger execution of a previously approved command without user interaction specific to that directory. User interaction is required to establish the initial approval, but subsequent replays occur silently. Exploitation requires low privileges on the system running OpenClaw and does not require network authentication beyond interacting with the approval surface.

For technical specifics, refer to the GitHub Security Advisory GHSA-3mq7-q27j-mq7q and the VulnCheck OpenClaw Advisory.

Detection Methods for CVE-2026-100530

Indicators of Compromise

  • Execution of previously approved commands from directories the user did not visit during the approval session
  • Approval records lacking a bound cwd or working-directory attribute
  • OpenClaw command execution against repositories or paths outside the originally reviewed project scope

Detection Strategies

  • Audit OpenClaw approval logs and correlate approved command signatures with the directories where they later executed
  • Flag any execution event whose working directory does not match the directory recorded at approval time
  • Compare installed OpenClaw versions against 2026.8.1 and alert on vulnerable builds in the fleet

Monitoring Recommendations

  • Forward OpenClaw audit and process-creation telemetry to a centralized log pipeline for retention and correlation
  • Alert on child processes spawned by OpenClaw that touch sensitive paths such as credential stores, CI configuration, or source repositories outside expected scopes
  • Track approval reuse frequency per command and surface anomalies when the same approval is invoked across many distinct directories

How to Mitigate CVE-2026-100530

Immediate Actions Required

  • Upgrade OpenClaw to version 2026.8.1 or later across all installations
  • Revoke existing allow-always approvals so users must re-grant them under the patched directory-aware logic
  • Inventory systems where OpenClaw runs and prioritize hosts with access to source code or production secrets

Patch Information

The fix is included in OpenClaw 2026.8.1. The patched release binds the working directory into the approval scope so reused approvals only apply to the directory in which they were originally granted. See the GitHub Security Advisory GHSA-3mq7-q27j-mq7q for release details.

Workarounds

  • Disable allow-always approvals and require per-invocation confirmation until patching is complete
  • Restrict OpenClaw to run only within explicitly scoped project directories using filesystem or container boundaries
  • Operate OpenClaw under a least-privileged account that cannot read sensitive repositories outside the intended workspace
bash
# Verify installed OpenClaw version and upgrade
openclaw --version
pip install --upgrade "openclaw>=2026.8.1"

# Clear persisted approvals so they are re-created under the patched policy
rm -f ~/.config/openclaw/approvals.json

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

Experience the Most Advanced Cybersecurity Platform

See how the world’s most intelligent, autonomous cybersecurity platform can protect your organization today and into the future.