A configuration injection issue in the Docker tool sandbox could allow dangerous Docker options (bind mounts, host networking, unconfined profiles) to be applied, enabling container escape or host data access.
openclaw (npm)<= 2026.2.14>= 2026.2.15 (next release)If an attacker can influence sandbox Docker configuration (or an operator pastes untrusted config), they may be able to:
- mount sensitive host paths (e.g. /etc, /proc, /sys, /dev, Docker socket)
- use network=host to bypass container network isolation
- use seccompProfile=unconfined / apparmorProfile=unconfined to weaken isolation
This can lead to host secret exfiltration or full host control (via Docker socket exposure).
OpenClaw now blocks dangerous sandbox Docker settings:
- runtime enforcement when building docker create args
- config-schema validation for network=host, seccompProfile=unconfined, apparmorProfile=unconfined
- security audit findings to surface dangerous sandbox docker config
agents.*.sandbox.docker.binds to mount system directories or Docker socket paths.agents.*.sandbox.docker.network at none (default) or bridge.unconfined for seccomp/AppArmor profiles.This advisory is pre-populated with the planned fixed version (>= 2026.2.15). Once openclaw@2026.2.15 is published to npm, publishing this advisory should be a single-click action.
Thanks @aether-ai-agent for reporting.
{
"github_reviewed_at": "2026-02-18T22:42:42Z",
"github_reviewed": true,
"cwe_ids": [
"CWE-250"
],
"nvd_published_at": "2026-02-20T00:16:16Z",
"severity": "HIGH"
}