CVE-2026-53813: OpenClaw: Fake package roots could influence memory-core artifact loading

When OpenClaw loads memory-core artifacts, it first finds the package root on the local disk. In affected versions, that lookup could be steered by workspace state, so a directory that imitates the package could be chosen instead of the bundled artifact root. This is an uncontrolled search path (CWE-427).

The attack is local. The attacker needs to put files into a workspace that OpenClaw uses, and the CVSS vector lists local access and low privileges. Artifacts loaded from the fake root run with the privileges of the OpenClaw process, so the vendor rates all three impact types as high. VulnCheck's write-up says impact could include running malicious code or reaching sensitive data.

The realistic exposure is a shared or untrusted workspace, such as a cloned repository or a directory other users can write to. A single-user machine running only trusted workspaces has little exposure.

Affected: npm package openclaw 2026.4.24 and earlier, when memory-core flows run in a workspace that someone else can influence.

npm ls openclaw

The flaw is in local package-root resolution, so it does not depend on network exposure of the Gateway.

Upgrade to openclaw 2026.4.25 or later, the first stable patched version.

npm install -g [email protected]

Before upgrading, run memory-core flows only from trusted workspaces. The advisory adds narrow channel and tool allowlists, and turning the feature off when it is not needed. It does not explain how the patch changes root resolution.

Search workspaces OpenClaw has used for directories that look like OpenClaw package roots, for example a stray package.json that names the package or memory-core files that are not in your install. Check for unexpected processes started by the OpenClaw process around memory-core use. The advisory names no log lines.

Summary

Fake package roots could influence memory-core artifact loading. In affected versions, a local package root resolution path influenced by workspace state could select a package root that was not the intended bundled artifact root.

This advisory is scoped to the named feature and configuration. It does not change OpenClaw's trusted-operator model: authenticated Gateway operators, installed plugins, and intentional local execution surfaces remain trusted unless a separate policy, approval, allowlist, sandbox, or auth boundary is crossed.

Impact

When the affected feature is enabled and reachable, this could load memory-core artifacts from an unintended local location. Practical impact depends on the operator's configuration and whether lower-trust input can reach that path.

Patched Versions

The first stable patched version is 2026.4.25.

Mitigations

run memory-core flows from trusted workspaces until patched. As general hardening, keep channel and tool allowlists narrow, avoid sharing one Gateway between mutually untrusted users, and disable the affected feature when it is not needed.