CVE Vulnerability Catalog

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.

OpenClaw Feishu before 2026.8.1 fails to validate whether a configured default account is disabled before selecting it for model tool operations. Attackers can exploit multi-account setups where a disabled default account retains credentials to read or modify Feishu resources through a revoked identity.

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.

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 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.

The @openclaw/slack npm package before 2026.8.1 contains an authorization flaw in its Slack download-file handler: when a file lacks the share metadata used to prove it belongs to the requested conversation, the conversation-authorization check fails open. An authenticated caller restricted to a single conversation who knows or obtains a file identifier can therefore download file contents from outside that conversation's scope, disclosing data across configured conversation boundaries.

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.

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.

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's Discord integration (npm package @openclaw/discord) before version 2026.9.3 could lose the sender-scoped media policy in the emoji and sticker upload actions before loading a local file. A sender permitted to invoke those actions could cause OpenClaw to read a host path that the same sender's configured media roots would otherwise reject, placing bytes from an out-of-policy local file into an outbound emoji or sticker upload.

The OpenClaw Prometheus diagnostics plugin (@openclaw/diagnostics-prometheus) before version 2026.9.3 does not enforce the operator.read scope on its authenticated metrics endpoint. In deployments using an identity-bearing Gateway authentication mode such as trusted-proxy, a caller whose effective role has no read scope can retrieve the diagnostics document even though ordinary read methods reject the same identity, disclosing operational metrics to an authenticated profile intentionally limited below read access.

Cotonti through version 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.

Cotonti through version 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 version 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 version 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.

Ghidra versions 9.2 through 12.1.4 contain a heap out-of-bounds read vulnerability in StringManager::getCodepoint when decoding multi-byte UTF-8, UTF-16, or UTF-32 characters without validating remaining buffer length. Attackers can craft malicious binaries containing strings or constant byte stores that end in multi-byte lead units to trigger out-of-bounds reads that crash the decompiler or leak adjacent heap memory into decompiled output.

Horilla before version 2.0.0 contains a CSRF vulnerability in the /attendance/approve-validate-attendance-request/ endpoint, where attendance state changes are performed via an HTTP GET before calling save(), so Django does not require CSRF validation. An unauthenticated attacker can cause a logged-in manager to make a top-level request that silently approves attendance with the victim's privileges and attributes the approval to the victim in the audit trail.

The convert.base64-encode, convert.quoted-printable-encode and convert.quoted-printable-decode stream filters accept a line-break-chars option whose length is tracked separately from the string itself. The filter constructors duplicate the value with pestrdup(), which stops at the first NUL byte, while keeping the original length. When the filter later emits a line break it copies the recorded length out of the truncated allocation, reading past its end and placing adjacent heap bytes into the filter output.

The IPv6 branch of the FastCGI client access check compares only the first 12 bytes of a 16-byte IPv6 address, so listen.allowed_clients matches on a /96 prefix instead of the exact address. An attacker who can source an address sharing the first 96 bits with an allowed one passes the check and reaches the FastCGI endpoint.

Horilla from version 1.0.0 to 1.6.0 and 2.0.0 contains a vulnerability in the get_mail_preview handlers in recruitment/views/actions.py and employee/not_in_out_dashboard.py, which render user-controlled content with the full request object in the Django template context. An authenticated user with a valid CSRF token can use template attribute traversal to read request.user.password, request.META, and related-user attributes, exposing password hashes, personal data, and server request metadata.