CVE-2026-28463: 
OpenClaw (formerly Moltbot or Clawdbot) vulnerability analysis and mitigation

Overview

CVE-2026-28463 is an arbitrary file read vulnerability in OpenClaw (npm package) affecting all versions prior to 2026.2.14. The flaw exists in the exec-approvals allowlist validation, which checks pre-expansion argv tokens but executes commands using a real shell (sh -c) that performs glob and environment variable expansion. This mismatch allows safe binaries such as head, tail, or grep to be weaponized to read arbitrary local files. It carries a CVSS v3.1 base score of 5.5 (Medium) and a CVSS v4.0 base score of 8.6 (High) (GitHub Advisory, Feedly).

Technical details

The root cause is classified as CWE-78 (Improper Neutralization of Special Elements used in an OS Command). OpenClaw's exec-approvals system validates argv tokens before shell expansion occurs, but then executes the command via sh -c, which expands glob patterns (e.g., *) and environment variables (e.g., $HOME/... or $FOO). Because the allowlist check sees only the unexpanded tokens, a safe bin like head or grep passes validation, but at execution time the shell expands the arguments into file paths, effectively turning a stdin-only binary into an arbitrary file reader. This vulnerability is only exploitable when tools.exec.host is set to gateway or node with security=allowlist; the default sandbox configuration is not affected (GitHub Advisory, Fix Commit).

Impact

Successful exploitation allows an authorized but untrusted caller, or an attacker leveraging prompt-injection, to cause the gateway or node process to disclose arbitrary files readable by that process. This can expose sensitive configuration data, credentials, private keys, or other confidential information stored on the host. The impact is limited to confidentiality (no integrity or availability impact under CVSS v3.1), but the scope of readable files is bounded only by the process's filesystem permissions (GitHub Advisory, Feedly).

Exploitability

There is no public proof-of-concept exploit and no evidence of in-the-wild exploitation at this time. The EPSS score is approximately 0.02%, reflecting low near-term exploitation probability. The vulnerability is not listed in the CISA Known Exploited Vulnerabilities catalog. Exploitation requires either an authorized caller with access to the exec tool or a successful prompt-injection attack against an AI agent using OpenClaw with host execution enabled in allowlist mode (Feedly, GitHub Advisory).

Exploitation steps

  1. Identify target configuration: Confirm the target OpenClaw instance is running with tools.exec.host set to gateway or node and security=allowlist, with one or more safe bins (e.g., head, tail, grep, wc) configured in tools.exec.safeBins.
  2. Craft a malicious command: Construct a command using a safe bin with a glob pattern or environment variable as an argument, e.g., head $FOO where FOO is set to a sensitive file path, or grep pattern * to match files in the working directory.
  3. Bypass allowlist validation: Submit the command through the exec tool. The allowlist checker sees head $FOO and validates $FOO as a non-path-like token, passing the check since head is a safe bin.
  4. Trigger shell expansion: The command is passed to sh -c, which expands $FOO to the actual file path (e.g., /etc/passwd or ~/.ssh/id_rsa), causing head to read and output the file contents.
  5. Exfiltrate output: The file contents are returned in the exec tool's output, disclosing sensitive data to the caller or the AI agent's context (GitHub Advisory, Fix Commit).

Indicators of compromise

  • Logs: Exec-approvals logs showing safe bin commands (e.g., head, tail, grep, wc) invoked with arguments containing $, *, or path-like tokens such as $HOME/, $FOO, or glob wildcards.
  • Process: Child processes spawned by the OpenClaw gateway or node process executing safe bins with expanded file paths as arguments (visible in process audit logs or auditd output).
  • File System: Unexpected access timestamps on sensitive files (e.g., /etc/passwd, SSH keys, .env files, configuration files) coinciding with OpenClaw exec activity.
  • Network: Outbound data transfers from the OpenClaw host following exec tool invocations, potentially indicating exfiltration of disclosed file contents.

Mitigation and workarounds

Upgrade OpenClaw to version 2026.2.14 or later, which fixes the issue by single-quoting all argv tokens at execution time via the new buildSafeShellCommand function, preventing glob and environment variable expansion for safe bin executions (Fix Commit). If immediate upgrade is not possible, disable host execution (tools.exec.host) or switch from security=allowlist to sandbox mode, as the default sandbox configuration is not affected. Additionally, review and minimize the tools.exec.safeBins list to only include binaries strictly necessary for operations, and implement monitoring for suspicious exec-approvals log entries involving environment variable or glob tokens (GitHub Advisory).

Community reactions

The vulnerability was reported by security researcher christos-eth and acknowledged by the OpenClaw maintainer (steipete) in the GitHub Security Advisory published February 15, 2026. The fix was released promptly in version 2026.2.14. The advisory was noted by automated security feeds including RedPacket Security and VulnCheck, which published their own advisory at vulncheck.com/advisories/openclaw-arbitrary-file-read-via-shell-expansion-in-safe-bins-allowlist. No significant broader media coverage or notable community debate has been identified (GitHub Advisory).

Additional resources


Source: This report was generated using AI

Related OpenClaw (formerly Moltbot or Clawdbot) vulnerabilities:

CVE ID

Severity

Score

Technologies

Component name

CISA KEV exploit

Has fix

Published date

CVE-2026-107392HIGH7.5
  • JavaScript logoJavaScript
  • openclaw
NoYesOct 08, 2026
CVE-2026-107391HIGH7.5
  • JavaScript logoJavaScript
  • openclaw
NoYesOct 08, 2026
CVE-2026-101899MEDIUM6.9
  • JavaScript logoJavaScript
  • commercial-gitlab-rails-ee-19.1
NoYesOct 13, 2026
GHSA-3x84-qq85-fj65MEDIUM6.3
  • OpenClaw (formerly Moltbot or Clawdbot) logoOpenClaw (formerly Moltbot or Clawdbot)
  • openclaw
NoYesOct 21, 2026
CVE-2026-107389MEDIUM6.2
  • JavaScript logoJavaScript
  • openclaw
NoYesOct 08, 2026

Free Vulnerability Assessment

Benchmark your Cloud Security Posture

Evaluate your cloud security practices across 9 security domains to benchmark your risk level and identify gaps in your defenses.

Request assessment

Additional Wiz resources

Get a personalized demo

Ready to see Wiz in action?

"Best User Experience I have ever seen, provides full visibility to cloud workloads."
David EstlickCISO
"Wiz provides a single pane of glass to see what is going on in our cloud environments."
Adam FletcherChief Security Officer
"We know that if Wiz identifies something as critical, it actually is."
Greg PoniatowskiHead of Threat and Vulnerability Management