OpenClaw (npm package openclaw) before 2026.8.1 could include deterministic hashes computed over the original, unredacted configuration in redacted configuration responses. When the Gateway password had low entropy and the remaining configuration values were reconstructable, these hashes acted as offline password verifiers: a caller able to obtain the redacted configuration (for example via config.get) could test password candidates offline without going through the rate-limited Gateway authentication path. Recovering the password could grant the documented shared-secret operator authority. Secret references were not affected in the same way. The issue is fixed in 2026.8.1.
OpenClaw (npm package 'openclaw') versions >= 2026.3.28 and < 2026.8.1 contain a credential exposure issue in memory embedding failover. When remote embedding fallback is configured and the primary embedding provider fails, the request can be sent to a different fallback provider while still reusing the primary provider's configured API key, causing that credential to be transmitted as a bearer token to an unintended vendor. The practical impact depends on the configured providers, whether failover occurs, and the privileges attached to the primary provider key. The issue is fixed in 2026.8.1; as a workaround, disable cross-provider embedding fallback or configure each provider with separate, narrowly scoped credentials.
OpenClaw (npm package 'openclaw') before 2026.8.1 fails to revoke memory tool access when an operator hot-disables memory configuration. Existing memory_search and memory_get tool instances retain the enabled configuration captured at creation time because the execution-time resolver treats explicit disablement like an unavailable configuration snapshot and restores the stale authority. As a result, during an already-running agent turn the model can continue searching and reading durable memory after the operator revoked that access, for the remainder of that run. Exploitation requires memory to be disabled while a previously created memory tool remains active. The issue is fixed in 2026.8.1.
OpenClaw versions before 2026.8.1 contain an authorization bypass vulnerability in webhook TaskFlow cancellation that allows attackers to cancel unrelated sessions. An attacker with a webhook route secret can supply an arbitrary child session key to cancel ACP or subagent work outside the route's configured authority.
OpenClaw (npm package `openclaw`) before 2026.8.1 incorrectly enforces sender tool policies during session-memory filename generation. In affected versions, filename generation created an embedded helper that retained tools which the originating sender's policy had removed. When session-memory filename generation was enabled for an agent reachable by lower-trust senders, model-mediated instructions could cause the helper to invoke tools outside that sender's effective policy; the demonstrated impact was the creation of persistent scheduled work. Exploitability depends on the model acting on the injected instruction and on which tools the helper exposes. The issue is fixed in 2026.8.1; as a workaround, disable session-memory filename generation for agents reachable by lower-trust senders.
OpenClaw versions before 2026.8.1 contain a path traversal vulnerability in QQBot voice attachment handling where filenames are decoded twice, allowing encoded traversal segments to reappear after sanitization. Attackers can supply crafted voice attachments that write files outside the intended staging directory to other process-writable locations.
OpenClaw (npm package `openclaw`) versions >= 2026.7.2 and < 2026.9.2 contain a race condition in the Discord realtime voice transcript path. Concurrent control-classified voice transcripts could consume speaker context belonging to another participant after an asynchronous control check, causing a transcript to inherit another speaker's owner status. In Discord agent-proxy voice sessions using the affected realtime control path, an utterance from a non-owner participant could reach the downstream agent boundary marked as owner, so owner-sensitive behavior is applied to the wrong speaker. Exploitation depends on concurrent transcript timing and on the tools and commands available to the affected agent. The issue is fixed in 2026.9.2; as a workaround, disable Discord realtime voice for agents that distinguish owner and non-owner senders.
OpenClaw (npm package 'openclaw') before 2026.8.1 does not apply the originating sender's global or per-agent toolsBySender policy when handling outbound attachments. A sender that has been explicitly denied filesystem read tools can still cause a known local file to be read and returned via a final-response media directive or a message attachment, disclosing local file contents to an admitted requester whose agent turn did not include the read tool. Exploitation requires knowledge or derivation of a useful host path and a delivery flow that accepts local attachments. The issue is fixed in version 2026.8.1.
OpenClaw (npm package 'openclaw') versions >= 2026.4.5 and < 2026.8.1 can lose the originating requester's restrictions and untrusted provenance when session-derived text is persisted to session memory. In deployments where session-memory capture and dreaming are enabled, a restricted external sender whose messages are admitted with limited tools can persist instructions that are later supplied to an unattended background (dreaming) agent holding broader file and command capabilities, allowing actions beyond the authority of the original turn and affecting files, commands, or services available to that agent. Exploitation requires the content to be captured, selected for later processing, and followed by the model. The issue is fixed in 2026.8.1.
OpenClaw versions before 2026.8.1 fail to validate all source fields in structured message attachments, allowing attackers to hide unvalidated host paths behind allowed attachment sources. Attackers can exploit this by providing multiple source fields to bypass sandbox path validation and cause Telegram delivery to read and send known host files that would otherwise be rejected.
OpenClaw before 2026.8.2 contains a denial of service vulnerability in the Browser extension relay that allows unauthenticated network sources to exhaust pending-authentication capacity. Attackers can hold every pending slot by maintaining silent WebSocket upgrades, preventing paired extensions from completing Browser Relay Authentication v2.
OpenClaw versions before 2026.8.1 contain an authorization scope widening vulnerability in file-transfer allow-always approvals that allows attackers to reuse standing grants for unreviewed paths. Attackers can exploit glob metacharacter interpretation and node display name reuse to access sibling paths or different nodes beyond the operator's original approval scope.
Cotonti through 1.0.0 contains a cross-site request forgery vulnerability in the extensions manager that allows attackers to perform state-changing actions without anti-CSRF token validation. Attackers can craft links or embed images to force administrators to install, update, pause, or unpause extensions by tricking them into visiting a malicious page while authenticated.
OpenClaw versions before 2026.8.1 contain a path traversal vulnerability in the tools.fs.workspaceOnly feature where Unicode filename fallback can normalize validated parent directory components. Admitted requesters can exploit canonically equivalent sibling directories to read files outside the configured workspace boundary.
OpenClaw versions before 2026.8.1 fail to bind working directory context to reusable exec approvals, allowing approved commands to execute in different directories. Attackers with an allow-always approval can reuse it to run the same command against unreviewed files or repositories with materially different effects.
OpenClaw (npm package 'openclaw') before 2026.8.1 could send third-party provider credentials to the wrong endpoint. In affected versions, when a third-party provider uses an OpenAI-compatible API and the resolved model metadata lacks a concrete base URL, a pinned session that continues after a model configuration hot reload retains that provider's credential while the OpenAI SDK selects its own default endpoint. A resulting request could disclose the configured third-party provider credential to an unrelated provider endpoint and fail with a misleading authentication error. Operators who observed this condition should rotate the affected credential. The issue is fixed in 2026.8.1.
Cotonti through 1.0.0 contains a reflected cross-site scripting vulnerability in message.php where the lng parameter is not properly escaped before output in the confirmation dialog. Unauthenticated attackers can craft malicious links with script payloads in the lng parameter to execute arbitrary JavaScript in victim browser sessions.
Cotonti through 1.0.0 contains an open redirect vulnerability in message.php that base64-decodes the redirect parameter without domain validation. Unauthenticated attackers can craft malicious links with encoded external URLs to redirect users to arbitrary sites via meta refresh tags for phishing attacks.
Cotonti through 1.0.0 contains a reflected cross-site scripting vulnerability in the search plugin highlight parameter that performs no HTML or JavaScript escaping. Attackers can craft malicious links with injected JavaScript in the highlight parameter that executes in the browser of any visitor who opens the link, including administrators.
OpenClaw Discord versions before 2026.7.1 have an authorization bypass vulnerability in guild metadata read actions. This flaw allows lower-trust users to access guild metadata that should be restricted by channel allowlists, bypassing configured read-target policies. The vulnerability has a medium severity rating with a CVSS score of 5.3.