OpenClaw CLI versions up to 2026.2.21-2 have a vulnerability in the system.run allowlist mode where shell-wrapper analysis can be bypassed using a shell line-continuation command substitution technique. This allows execution of non-allowlisted subcommands despite the allowlist restrictions. The issue affects deployments using tools.exec.security=allowlist with ask=on-miss or off. A fix is planned in version 2026.2.22. Temporary mitigations include setting tools.exec.ask=always or tools.exec.security=deny.
OpenClaw CLI versions up to 2026.2.26 have an unauthenticated denial-of-service vulnerability in the Zalo webhook endpoint. Attackers can cause unbounded memory growth by sending requests with varying query-string keys to the same webhook path, leading to potential process instability or out-of-memory conditions. This issue is fixed in version 2026.3.1 by normalizing keys and bounding the tracking state to prevent unbounded growth.
OpenClaw CLI versions prior to 2026.2.13 contain a path traversal vulnerability in the browser control API. This flaw allows user-supplied output paths for trace and download files to escape the intended temporary directories, potentially enabling arbitrary file writes depending on filesystem permissions. The issue affects versions from 2026.1.29 up to but not including 2026.2.13. The vulnerability has been fixed in version 2026.2.13 and later by enforcing strict path constraints on relevant API endpoints.
OpenClaw CLI versions before 2026.2.14 have a vulnerability in the exec-approvals allowlist feature where safe binaries intended to be stdin-only can bypass restrictions via shell expansion. This allows authorized but untrusted users to cause the gateway or node process to read arbitrary local files when host execution is enabled in allowlist mode. The issue arises because the allowlist validation checks tokens before shell expansion, but execution uses a real shell that expands globs and environment variables. A fix is available in version 2026.2.14 that forces argv tokens to be treated as literal text, preventing this bypass.
OpenClaw CLI versions prior to 2026.2.12 have a vulnerability in the hooks authentication mechanism where token comparison is not performed in constant time. This timing side-channel flaw could allow an attacker with network access to the hooks endpoint to infer the authentication token by measuring response times over many requests. The vulnerability requires the hooks endpoint to be exposed to an untrusted network and is complicated by real-world network latency and jitter. The issue is fixed in version 2026.2.12 and later.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.2.14 contain a critical remote code execution vulnerability (CVE-2026-28466) in the gateway-to-node invocation path. An authenticated gateway client could bypass node exec approval controls by injecting internal approval fields into the node.invoke parameters, enabling arbitrary command execution on connected node hosts. This can lead to full compromise of developer workstations, CI runners, and servers running the node host. The vulnerability is patched in OpenClaw version 2026.2.14.
OpenClaw CLI versions prior to 2026.2.2 are vulnerable to a server-side request forgery (SSRF) via attachment or media URL hydration. An attacker able to control the media URL can cause the gateway to fetch arbitrary HTTP(S) URLs, including internal network endpoints, and send the fetched content as an outbound attachment. This vulnerability allows data exfiltration from internal or cloud metadata services but does not enable code execution or system compromise. The issue is fixed in version 2026.2.2 by implementing SSRF protections such as blocking private IP ranges, DNS pinning, and redirect handling.
OpenClaw CLI versions from 2026.1.29-beta.1 up to but not including 2026.2.14 contain an authentication bypass vulnerability in the sandbox browser bridge server. This local-only vulnerability allows any process on the same machine to access the bridge server without authentication, enabling full browser session compromise including tab enumeration, navigation, JavaScript execution, and data exfiltration. The issue is patched in version 2026.2.14. Users should upgrade or disable the sandboxed browser to mitigate the risk.
OpenClaw CLI versions prior to 2026.2.14 have a vulnerability in Google Chat webhook handling where multiple webhook targets registered on the same HTTP path can cause cross-account policy misrouting. This occurs because the system routes inbound webhook events to the first matching target that verifies the request, potentially processing events under the wrong account context. The issue is fixed in OpenClaw version 2026.2.14. Users of the deprecated clawdbot package should migrate to OpenClaw and upgrade accordingly. A workaround is to ensure each webhook target uses a unique path to avoid ambiguous routing.
OpenClaw CLI versions from 2026.1.14-1 up to but not including 2026.2.2 have a vulnerability in the Matrix DM allowlist matching logic. The allowlist can be bypassed due to identity confusion caused by matching on attacker-controlled display names and localparts of Matrix IDs without homeserver parts. This allows a remote Matrix user to impersonate an allowed identity and bypass allowlist restrictions. The issue affects deployments with the optional Matrix plugin enabled. A fix is available in version 2026.2.2 and later.
OpenClaw CLI versions up to 2026.2.1 have a vulnerability in the gateway WebSocket connect handshake that allows skipping device identity checks if an auth.token is present but not validated. This could enable unauthorized clients to connect without proper device identity or pairing, potentially granting operator access depending on configuration. The issue is fixed in version 2026.2.2 by requiring validated shared-secret authentication for skipping device identity. The gateway should be restricted to trusted networks and users to mitigate risk.
OpenClaw CLI versions prior to 2026.2.2 have an authorization bypass vulnerability where clients with only 'operator.write' permissions can approve or deny exec requests by sending the '/approve' chat command. This bypasses intended permission scoping, allowing unauthorized approval actions. The issue mainly affects multi-client setups with scoped tokens. A fix is available in version 2026.2.2 that enforces proper permission checks for the '/approve' command.
The openclaw-cli package is affected by a server-side request forgery (SSRF) vulnerability in its optional Tlon (Urbit) extension authentication. This vulnerability arises because the extension accepted a user-provided base URL for authentication and used it to make outbound HTTP requests without sufficient validation. Only deployments that have installed and configured the Tlon extension and allow attackers to influence the Urbit URL are impacted. The issue is fixed in version 2026.2.14, which introduces URL validation and an SSRF guard blocking private/internal hosts by default.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.2.14 have a vulnerability in the manual Chutes OAuth login flow. This flaw allows bypassing OAuth CSRF state validation if an attacker convinces a user to paste attacker-controlled OAuth callback data. As a result, the attacker can cause OpenClaw to exchange an authorization code and persist tokens for the wrong Chutes account. The automatic local callback flow is not affected. A fix is available that enforces full redirect URL validation including state verification.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.2.13 contain a vulnerability where multiple webhook handlers accept and buffer request bodies without strict unified byte or time limits. This allows a remote unauthenticated attacker to send oversized payloads, causing memory pressure and degrading service availability, resulting in a denial of service (DoS). A patch is available that adds bounded request-body handling and stream guards to mitigate this issue.
OpenClaw CLI versions up to 2026.2.14 used the deprecated SHA-1 hashing algorithm for sandbox identifier cache keys. This hashing was used to determine if existing sandbox containers could be safely reused. Because SHA-1 is vulnerable to collisions, this posed a risk of cache poisoning and unsafe sandbox state reuse. The vulnerability is fixed in version 2026.2.15 by replacing SHA-1 with SHA-256 for these hashes, restoring collision resistance.
OpenClaw CLI versions before 2026.2.14 allowed Telegram allowlist authorization based on mutable usernames instead of immutable numeric sender IDs. This flaw could let unauthorized users gain access if a username is recycled or changed hands. The vulnerability has been fixed by requiring numeric Telegram sender IDs only and rejecting username-based allowlist entries. A security audit warning and a fix command are provided to help migrate legacy configurations.
OpenClaw CLI versions up to 2026.1.30 have a vulnerability in the MS Teams extension where bearer tokens used for authorization can be leaked to allowlisted suffix domains during attachment downloads. This occurs because the default allowlist uses suffix matching, potentially sending tokens to untrusted but allowlisted hosts. The issue is fixed in version 2026.2.1. Users not using MS Teams are not affected. Mitigation includes upgrading to 2026.2.1 or disabling the MS Teams extension if not needed.
OpenClaw CLI versions up to 2026.2.9 have a path traversal vulnerability due to unsanitized session IDs and session file paths used in transcript file operations. An authenticated attacker with access to the gateway can exploit this to read or write files outside the intended sessions directory. The vulnerability is fixed in version 2026.2.12 by validating session IDs and enforcing directory containment.
OpenClaw CLI versions from 2026.1.16-2 up to but not including 2026.2.14 contain a path traversal (Zip Slip) vulnerability in archive extraction during explicit installation commands. This flaw allows crafted archives to write files outside the intended extraction directory, potentially enabling arbitrary file writes as the current user. The vulnerability affects commands such as 'openclaw skills install', 'hooks install', 'plugins install', and 'signal install' when run against untrusted archives. It does not affect normal gateway operations or message receipt. The issue is fixed in version 2026.2.14.