NewApp Monitoring is liveSign up free
AgentbeamBeta
AI agent vulnerability · OpenClaw

CVE-2026-43534: OpenClaw: Agent hook events could enqueue trusted system events from unsanitized external input

mediumCVSS 4.0 6.6 · 3.1 9.1GHSA-7g8c-cfr3-vqqr

OpenClaw before 2026.4.10 turned externally supplied agent hook metadata into trusted system events. Untrusted input entered the agent as higher-trust context. Upgrade the `openclaw` npm package to 2026.4.10 or later, where hook names are sanitized and hook system events are marked untrusted.

Published Apr 17, 2026 · Updated May 8, 2026 · Analysis Oct 29, 2026

What an attacker can do

OpenClaw dispatches agent hooks and queues system events for the agent. System events carry more trust than ordinary user messages. In the affected versions, the hook name and metadata came from outside, were not sanitized, and the resulting event was enqueued as trusted (CWE-345, CWE-269). A party who could supply hook metadata could therefore write text that the agent treated as a system-level message.

That is a prompt-injection path with elevated trust. The agent may follow instructions from the event that it would refuse from a normal sender. What the injected event can do depends on the deployment's tool policy, and the advisory gives no proof of concept.

The CVSS 3.1 vector shows a network attack, no privileges, no user interaction, and high confidentiality and integrity impact. The CVSS 4.0 vector is lower and carries an unproven exploit maturity flag. Deployments that accept hook calls from outside the host, such as webhooks, are the exposed ones.

Who is affected

The openclaw npm package, all versions below 2026.4.10. The exposure is larger where agent hook endpoints can be reached by parties you do not fully trust. Check the installed version:

npm ls -g openclaw

openclaw

Ecosystem
npm
Affected
< 2026.4.10
Fixed in
2026.4.10
Check
npm ls openclaw

How to fix it

Upgrade to 2026.4.10 or newer. The advisory says 2026.4.14 includes the fix.

npm install -g openclaw@latest

The fix (PR #64372, commit e3a845b) sanitizes hook names and marks agent hook system events as untrusted before enqueueing. The advisory lists no workaround. Until you upgrade, restrict who can reach your hook endpoints; that is my suggestion, not the advisory's.

How to tell if you were hit

Inspect the agent's event queue and session history for system events whose source is an agent hook, especially hook names with unusual characters or long text. Check hook endpoint access logs for callers you do not recognize. Look for tool calls that follow a hook event and were not requested by a user. The advisory gives no log signature.

Original advisory

From the GitHub Advisory Database (CC BY 4.0).

Summary

Agent hook events could enqueue trusted system events from unsanitized external input.

Affected Packages / Versions

  • Package: openclaw
  • Ecosystem: npm
  • Affected versions: < 2026.4.10
  • Patched versions: >= 2026.4.10

Impact

Agent hook dispatch could turn externally supplied hook metadata into trusted system events, allowing untrusted input to enter the agent as higher-trust context.

Technical Details

The fix sanitizes hook names and marks agent hook system events as untrusted before enqueueing them.

Fix

The issue was fixed in #64372. The first stable tag containing the fix is v2026.4.10, and [email protected] includes the fix.

Fix Commit(s)

  • e3a845bde5b54f4f1e742d0a51ba9860f9619b29
  • PR: #64372

Release Process Note

Users should upgrade to openclaw 2026.4.10 or newer. The latest npm release, 2026.4.14, already includes the fix.

Credits

Thanks to @zsxsoft, with sponsorship from @KeenSecurityLab and @qclawer for reporting this issue.

Credited to zsxsoft, qclawer, KeenSecurityLab.

Questions

Which OpenClaw version fixes CVE-2026-43534?

2026.4.10. All earlier versions are affected, and 2026.4.14 includes the fix.

What does the flaw let an attacker do?

Put untrusted input into the agent as a trusted system event, so it is treated with higher trust than a normal message.

Why do the CVSS scores differ so much?

The 3.1 vector scores 9.1 with high confidentiality and integrity impact. The 4.0 vector scores 6.6 and notes unproven exploit maturity. Impact in practice depends on your tool policy.

Is it exploited in the wild?

No source I read says so, and the entry has no CISA KEV data.

References

How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. The analysis is written by an AI model from the sources listed above and checked by automated rules before it is published. Check versions against the vendor's advisory before you act. Methodology

More OpenClaw vulnerabilities

All OpenClaw entries