A vulnerability in openclaw-cli versions before 2026.4.12 allows empty approver lists to grant explicit approval authorization. This could enable an unauthorized sender to resolve pending approvals if they know an approval ID. The issue is fixed in version 2026.4.12 and later.
OpenClaw CLI versions before 2026.4.10 have a vulnerability where the sandbox browser CDP relay could expose the Chrome DevTools Protocol on all network interfaces (0.0.0.0), potentially allowing unintended external access. This exposure occurs because the CDP relay binds too broadly instead of restricting access to local or sandboxed sources. The issue is fixed in version 2026.4.10 and later, which enforce source-range restrictions by default to prevent broad exposure.
OpenClaw CLI versions before 2026.4.22 contain a vulnerability where the Zalo plugin could forward attacker-controlled outbound photo URLs to the Zalo Bot API without proper SSRF validation. This issue was fixed in version 2026.4.22 by enforcing SSRF hostname validation on outbound photo URLs.
OpenClaw CLI versions before 2026.4.20 have a vulnerability where hook mapping templates could bypass the intended session-key opt-in control. This flaw allows a hook mapping to produce an externally influenced session key even when the configuration to disallow request session keys is enabled. The issue affects webhook routing isolation but does not enable host execution by itself. The vulnerability is fixed in version 2026.4.20.
Mattermost slash token revocation in OpenClaw CLI could lag until the monitor refreshes, allowing brief acceptance of revoked tokens. This affects versions from 2026.1.29 up to but not including 2026.4.24. The issue occurs only when the affected feature is enabled and reachable, with impact depending on operator configuration and trust boundaries. A patch is available in version 2026.4.24. Mitigation includes restarting or refreshing the Mattermost monitor after token rotation until patched, and applying configuration hardening such as narrow allowlists and disabling the feature if not needed.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.4.22 have a vulnerability where Slack and Zalo webhook secrets may remain active after a secrets.reload operation. This allows a caller with an old webhook secret to continue sending webhook events during a brief stale-secret window after revocation. The issue affects only the named feature and does not alter the trusted-operator security model. The vulnerability has a medium severity rating with a CVSS score of 6.5. A patch is available starting from version 2026.4.22. Operators are advised to restart the affected channel runtime after rotating webhook secrets as a mitigation until they apply the patch.
A vulnerability in OpenClaw CLI's Feishu dynamic-agent bindings allows sender-created bindings to bypass configured config-write enforcement. This can lead to changes in sender-agent binding state beyond intended policy when the feature is enabled and reachable. The issue affects versions from 2026.1.29 up to but not including 2026.5.6. The vulnerability has a low severity score and does not affect confidentiality or availability, only integrity to a limited extent. A patch is available in version 2026.5.6. Mitigations include disabling sender-created Feishu dynamic-agent bindings if not needed and restricting channel and tool allowlists.
A vulnerability in openclaw-cli allows a workspace .env file to influence the npm_execpath used during bundled runtime dependency installation. This can cause an unintended local package-manager executable to run if a trusted operator opens a repository containing such a .env file. The issue affects versions from 2026.1.29 up to but not including 2026.4.29. A patch is available in version 2026.4.29. The impact depends on the operator's configuration and trust boundaries. Mitigations include installing dependencies only from trusted workspaces and restricting access via allowlists and disabling the affected feature when not needed.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.4.25 contain a vulnerability where internal or webchat command authorization can inherit a wildcard ownerAllowFrom state across channel boundaries. This may allow owner-style command behavior to be executed beyond its intended channel scope. The impact depends on the operator's configuration and whether lower-trust inputs can access the affected paths. A patch is available in version 2026.4.25. Until patched, explicit per-channel owner command allowlists and narrow allowlists are recommended, along with disabling the affected feature if not needed.
OpenClaw CLI versions from 2026.2.13 up to 2026.3.1 allowed node camera URL payloads to be fetched by the gateway or agent host without restricting downloads to the resolved node host. This could enable a malicious or compromised paired node to cause the gateway to fetch URLs from off-node destinations, potentially facilitating internal network probing or fetch pivots. The issue is considered medium severity and has been fixed in version 2026.3.2 by enforcing hostname matching and SSRF-guarded fetches.
OpenClaw CLI versions prior to 2026.2.23 had a vulnerability in the command authorization logic where conversation identifiers in the allowlist could authorize unintended participants. This occurred because the system treated conversation identities as valid senders, broadening command execution permissions beyond the intended sender. The issue was fixed in version 2026.2.23 by restricting authorization checks to sender-only identities and excluding conversation identifiers except as a fallback in direct-message contexts.
OpenClaw CLI had an issue where authenticated non-owner users in direct message sessions could invoke certain owner-only gateway actions through a narrow tool invocation path. This vulnerability does not allow unauthenticated access or direct code execution. The root cause was inconsistent enforcement of owner-only access controls, relying on tool-name checks rather than comprehensive metadata-driven scopes. The issue affects openclaw versions up to 2026.2.17 and was fixed in version 2026.2.19 by refactoring scope mappings and centralizing owner-only enforcement.
OpenClaw CLI versions up to 2026.2.17 are affected by a vulnerability in the handling of remote iMessage attachments fetched via SCP. The issue stems from using trust-on-first-use host-key behavior and lack of strict validation of remote host tokens, which could allow a man-in-the-middle or DNS poisoning attack on first connection. The vulnerability is mitigated by enforcing strict SSH host key checking, validating remote host tokens, and adding argument barriers to SCP commands. A fix is available in versions 2026.2.19 and later.
OpenClaw CLI has a vulnerability where Unicode canonicalization drift in node metadata policy classification can cause a paired node to bypass intended command restrictions. This allows the paired node to supply Unicode-confusable platform or deviceFamily metadata that passes metadata pinning but is classified differently during command policy resolution, broadening node command allowlists beyond intended defaults. The issue affects versions up to 2026.2.26 and is fixed in version 2026.3.1 by hardening metadata canonicalization and making unknown platform defaults more restrictive.
OpenClaw CLI's system.run allowlist mode did not properly detect PowerShell encoded-command wrappers such as -EncodedCommand, -enc, or -e. This allowed callers with access to system.run to execute PowerShell inline payloads without the usual approval step required for equivalent -Command invocations. The issue was fixed in version 2026.3.7 by updating the shell-wrapper parsing to recognize these encoded-command aliases, ensuring allowlist mode enforces approval consistently.
OpenClaw CLI versions up to 2026.2.24 have a vulnerability where hardlink aliases inside a workspace can reference files outside the workspace boundary in certain workspace-only restricted configurations. This allows reading and modification of out-of-workspace files through these hardlink aliases. The issue affects deployments that enable workspace-only filesystem restrictions. A fix is available in version 2026.2.25, which rejects hardlinked final-file aliases during workspace boundary validation.
A vulnerability in the Openclaw CLI allowed sandboxed sessions using runtime="acp" to bypass sandbox inheritance and initialize host-side ACP runtime. This sandbox-boundary bypass could lead to high impact on confidentiality, integrity, and availability. The issue affects Openclaw versions from 2026.1.29 up to but not including 2026.3.2. A patch is available in version 2026.3.2 that denies ACP spawn from sandboxed requester sessions and rejects sandbox="require" for runtime="acp".
OpenClaw CLI versions up to 2026.2.21-2 have an authorization bypass vulnerability in Discord allowlists due to slug normalization collisions of user name/tags. This can cause unintended access when different Discord user tags normalize to the same slug, bypassing name-based allowlist checks. The issue is fixed in version 2026.2.22 by preferring resolved user IDs for authorization. Users are advised to prefer stable Discord user IDs in allowlists and run the security audit tool to identify and address warnings.
A vulnerability in the OpenClaw CLI's Windows system.run approval process allows trailing arguments after 'cmd.exe /c' to execute without proper approval or audit text reflection. This mismatch can lead to local command execution on a trusted Windows node under an authenticated operator context. The issue affects OpenClaw versions up to 2026.2.19-2 and is planned to be fixed in version 2026.2.21.
OpenClaw CLI versions up to 2026.2.21-2 have a vulnerability where adding 'sort' to the non-default safeBins configuration allows bypassing intended allowlist approval constraints via the --compress-program option. This can invoke an external helper, circumventing security controls. A patch is planned for version 2026.2.22.