A vulnerability in OpenClaw CLI's Discord allowFrom feature allows binding to mutable display names. This can cause a Discord account that can change display or global name metadata to match a policy entry incorrectly. The issue affects versions from 2026.1.29 up to but not including 2026.5.7. The vulnerability could allow an attacker to receive agent access intended for another Discord identity, depending on configuration. A patch is available in version 2026.5.7. Until patched, using stable Discord user IDs in allowlists is recommended, along with narrowing allowlists and disabling the feature if not needed.
The OpenClaw CLI 'focus' command could bypass controlScope enforcement in affected versions, allowing the command to run without enforcing the expected control scope. This vulnerability affects versions from 2026.1.29 up to but not including 2026.4.25. The impact depends on the operator's configuration and whether lower-trust inputs can reach the vulnerable command path. The issue does not alter OpenClaw's trusted-operator model, meaning trusted Gateway operators and installed plugins remain trusted unless other security boundaries are crossed. A fix is available in version 2026.4.25. Until patched, restricting access to the focus command to trusted operators is recommended, along with narrowing allowlists and disabling the feature if not needed.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.5.12 have a vulnerability where Slack reaction events could bypass reaction notification settings, causing these events to be processed even when disabled. This issue affects the Slack reaction event handling feature and does not alter the trusted-operator security model of OpenClaw. The vulnerability may lead to unintended agent processing depending on configuration and trust boundaries.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.4.25 contain a vulnerability where an empty-scope device re-pairing request can bypass intended scope containment controls. This may allow broader access scopes than intended during device re-pairing. The vulnerability affects configurations where the feature is enabled and reachable, with impact depending on operator setup and input trust levels. A patch is available in version 2026.4.25. Mitigations include revoking unexpected device sessions, requiring fresh pairing for suspicious devices, narrowing allowlists, avoiding shared Gateways between untrusted users, and disabling the affected feature if not needed.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.5.12 on Linux and macOS have a vulnerability where exec allowlist entries using argPattern to restrict executable arguments are bypassed. This allows disallowed arguments to be executed without approval prompts, potentially enabling unintended file, network, or command access. The issue does not affect Windows or path-only allowlist entries. A fix is available in version 2026.5.12.
A vulnerability in openclaw-cli allows shell positional parameters to weaken strict inline-eval checks. This can enable shell-provided content to execute outside the intended allowlist rules when the affected feature is enabled. The issue affects versions from 2026.1.29 up to but not including 2026.4.2. The vulnerability has a high severity with a CVSS score of 8.1. A patch is available in version 2026.4.2. Mitigations include avoiding allowlisting shell carrier patterns and requiring approval for shell wrappers until patched.
A vulnerability in openclaw-cli's config recovery feature could restore the openclaw.json configuration file with overly broad file permissions. This may expose sensitive local configuration data to other users on the same host if OS permissions allow. The issue affects versions from 2026.4.23 up to but not including 2026.4.24. The vulnerability does not alter the trusted-operator model of OpenClaw but could increase local exposure when the feature is enabled. A fix is available in version 2026.4.24.
A vulnerability in OpenClaw's Zalo allowFrom feature allows binding to mutable display names, potentially causing agent responses to be misdirected to another Zalo identity. This affects versions from 2026.1.29 up to but not including 2026.5.4. The issue depends on operator configuration and the trust level of inputs reaching the affected feature. A fix is available starting with version 2026.5.3. Mitigations include using stable Zalo identifiers, restricting friend access, narrowing allowlists, avoiding shared Gateways between untrusted users, and disabling the feature if not needed.
A vulnerability in openclaw-cli allows a workspace .env file to influence the STATE_DIRECTORY environment variable, potentially causing bundled runtime dependencies to be loaded from unintended local paths. This affects versions from 2026.1.29 up to but not including 2026.5.2. The issue requires the affected feature to be enabled and reachable and depends on operator configuration and trust boundaries. The vulnerability does not alter the trusted-operator model of OpenClaw. A patch is available in version 2026.5.2. Mitigations include avoiding opening untrusted workspace .env files before patching and restricting access via allowlists and disabling the feature if not needed.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.5.28 have a vulnerability where hostname checks treat trailing-dot hosts inconsistently. This can allow requests with trailing-dot hostnames to bypass blocklist comparisons, potentially reaching destinations that operators intended to block. The vulnerability affects features that accept model- or workspace-derived URLs and depends on operator configuration and input trust levels. A patch is available starting with version 2026.5.26. Until patched, network-level blocking of private-network and metadata destinations is recommended, along with minimizing allowlists and disabling the affected feature if not needed.
BlueBubbles sender policy in OpenClaw CLI versions from 2026.1.29 up to but not including 2026.5.7 could match mutable conversation identifiers instead of stable sender identities. This allows a participant who can influence conversation-level identifiers to bypass intended sender restrictions. The vulnerability affects configurations where the feature is enabled and reachable, potentially allowing agent responses to be received by unintended senders. The issue does not affect the trusted-operator model of OpenClaw. A patch is available in version 2026.5.7. Mitigations include preferring stable sender identifiers, restricting BlueBubbles groups, narrowing allowlists, avoiding shared Gateways between untrusted users, and disabling the affected feature if not needed.
A vulnerability in OpenClaw CLI on macOS allows combined POSIX inline-command flags to bypass the exec allowlist, potentially executing shell content outside intended restrictions. This affects versions from 2026.1.29 up to but not including 2026.5.6. The impact depends on the operator's configuration and whether untrusted input can reach the vulnerable code path. A patch is available in version 2026.5.6. Until patched, requiring approval for combined shell flag forms on macOS is recommended, along with general hardening such as narrowing allowlists and disabling the feature if not needed.
A vulnerability in openclaw-cli allows replay of bootstrap tokens to widen pending pairing scopes. An attacker with access to a pending bootstrap token can reuse it before approval to request broader scopes than originally intended. This issue affects versions from 2026.1.29 up to but not including 2026.5.12. The vulnerability depends on the operator's configuration and whether untrusted inputs can reach the affected feature. A fix is available in version 2026.5.12.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.4.25 contain a vulnerability where tool group policy callers could accept unvalidated group IDs. This may cause the application of incorrect group-policy decisions for tool invocations when the affected feature is enabled and accessible. The vulnerability does not alter the trusted-operator model of OpenClaw. A patch is available starting with version 2026.4.25. Until patched, it is recommended to avoid exposing group-policy controlled tools to untrusted senders and to apply configuration hardening measures.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.5.28 have a vulnerability where the host environment sanitizer misses two Node.js control variables. This allows lower-trust environment sources, such as workspace .env files or tool overrides, to pass these control variables through the sanitizer. The issue affects the environment of Node.js child processes launched under certain configurations, potentially influencing their behavior. The vulnerability does not alter the trusted-operator model of OpenClaw. A fix is available starting with version 2026.5.26. Mitigations include avoiding untrusted environment inheritance and restricting access through allowlists and disabling the affected feature if not needed.
A vulnerability in openclaw-cli allows a workspace-derived service PATH to influence the selection of the trash command during maintenance tasks. This could cause an unintended local executable to run if the operator's configuration allows lower-trust input to affect the PATH. The issue affects versions from 2026.1.29 up to but not including 2026.5.2. The vulnerability does not alter the trusted-operator model of OpenClaw. A fix is available in version 2026.5.2.
OpenClaw CLI versions from 2026.1.29 up to but not including 2026.5.12 have a vulnerability in shell inline-command parsing where an allowlist check can be bypassed. This allows shell commands to run without the intended approval or allowlist prompt if the affected feature is enabled and reachable. The vulnerability does not affect the trusted-operator model but depends on the operator's configuration and input trust levels. A patch is available in version 2026.5.12.
OpenClaw CLI's Zalo plugin had a missing authorization check for group senders, allowing unauthorized GROUP messages to bypass intended allowlist restrictions. This flaw permitted senders not on the configured group allowlist to trigger agent dispatch processing. The issue was fixed in version 2026.2.24 by adding explicit runtime group-policy evaluation and fail-closed behavior for missing configurations.
OpenClaw CLI versions up to 2026.2.1 contain a server-side request forgery (SSRF) vulnerability in the Image tool's remote fetch functionality. This flaw allows attackers to make HTTP requests to internal or restricted network targets by supplying crafted URLs. The vulnerability is patched in version 2026.2.2 and later. Exploitation requires attacker control over the Image tool invocation and is limited by media-type validation and request constraints. The SSRF can enable internal network probing and access to unauthenticated internal HTTP endpoints.
OpenClaw CLI has a vulnerability in its hook authentication rate limiter where IPv4 and IPv4-mapped IPv6 client addresses are treated as distinct, allowing attackers to bypass rate limits. This flaw enables an attacker to double the number of allowed failed authentication attempts within a 60-second window by splitting attempts across both address forms. The issue affects versions up to and including 2026.2.21-2 and is fixed in the planned release 2026.2.22.