Disclosed weakness · no CVE recorded

OpenClaw sandboxed web searches could use a provider outside the configured trust policy

OpenClaw · OpenClaw sandboxed web search

First documented disclosure · Review checkpoint

Summary

A sandboxed agent’s search could run through a configured third-party provider even when its sandbox policy allowed only bundled or verified official providers.

Severity and exploitation evidence

No CVSS score recorded in this review.

Exploitation not established: The project documented the defect and its repair. The reviewed sources identify no malicious use.

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The search provider selection step lost the sandbox state and skipped its trust filter.

Conditions required

A sandboxed agent uses web search with a configured provider excluded by the sandbox trust policy.

Potential impact

A provider outside the intended trust boundary could execute the search and receive its inputs.

Affected versions and configurations

The project has not published a complete affected version range.

Fix and mitigation status

OpenClaw 2026.11.2 includes the fix. The exact first fixed release across all release lines is unverified.

Practical next steps

Update OpenClaw and select an eligible search provider for sandboxed agents.

Operating lesson

Apply the sandbox policy when choosing the provider that will actually run the request.

Evidence and limits

The original September issue concerns web search. The fix preserves existing provider selection for agents outside the sandbox.

Sources