Executive brief
OpenClaw, a tool for managing automated skills and commands, contains a flaw where certain commands can bypass security checks. An attacker with basic access could execute specific commands that skip the system's auditing and policy enforcement hooks. This could allow unauthorized actions to occur without being logged or blocked by established security rules.
Technical details
OpenClaw versions prior to 2026.5.6 are vulnerable to a hook bypass (CWE-693) in the skill-command dispatch path. The root cause is that certain dispatch paths fail to trigger the 'before-tool-call' hook, which is responsible for auditing and policy enforcement. A remote attacker with low privileges can exploit this by routing skill commands through the affected path to execute tools without triggering security middleware. This bypasses parity with other tool entry points that are correctly covered by hooks. The issue is resolved in version 2026.5.6.
Affected products
- OpenClaw openclaw <= 2026.5.5
Timeline
- 2026-05-28: advisory: Original GHSA-68xw-r643-9p5w published
- 2026-06-16: disclosed: NVD and VulnCheck publication of CVE-2026-53845
- 2026-06-18: patched: Duplicate advisory GHSA-r7vv-6763-m739 withdrawn in favor of original