CVE Feed

pnpm is a package manager. Prior to 10.34.2 and 11.5.3, pnpm can persist package-manager bootstrap metadata in the first YAML document of pnpm-lock.yaml. Before the patch, direct pnpm execution trusted an already resolved packageManagerDependencies entry when the committed env lockfile contained matching pnpm and @pnpm/exe versions. A malicious repository could therefore commit package-manager lockfile package records and snapshots that bypassed fresh package-manager resolution, then cause pnpm to install and execute bytes selected by that committed lockfile state during automatic version switching. This vulnerability is fixed in 10.34.2 and 11.5.3.

pnpm is a package manager. Prior to 10.34.2 and 11.5.3, pnpm can install configDependencies declared in pnpm-workspace.yaml before command dispatch. Before the patch, a repository could declare pacquet or @pnpm/pacquet as a config dependency and pnpm treated that repository-controlled dependency as an install-engine opt-in. During install, pnpm resolved a platform-specific @pacquet/<platform>-<arch>/pacquet binary from node_modules/.pnpm-config/<packageName> and spawned it as the developer or CI user. This vulnerability is fixed in 10.34.2 and 11.5.3.

pnpm is a package manager. Prior to 10.34.2 and 11.5.3, the generic peer-suffix normalizer also stripped parenthesized text from git, URL, tarball, file, and other opaque locators. Approval for one source string could therefore authorize a different attacker-controlled source whose locator normalized to the same value. This vulnerability is fixed in 10.34.2 and 11.5.3.

pnpm is a package manager. Prior to 10.34.2 and 11.5.3, pnpm and pacquet expanded ${ENV_VAR} placeholders from repository-controlled .npmrc and pnpm-workspace.yaml into registry request destinations and registry credentials. A malicious repository could cause dependency resolution to send victim environment secrets to an attacker-selected registry before lifecycle scripts run. This vulnerability is fixed in 10.34.2 and 11.5.3.

jq is a command-line JSON processor. Prior to 1.8.2, on 32bit system, jvp_string_append has a chance of integer/multiple overflowing and then causing a massive buffer overrun. This vulnerability is fixed in 1.8.2.

pnpm is a package manager. Prior to 10.34.0 and 11.4.0, `pnpm install` in non-frozen mode can accept new remote package content after detecting that the downloaded tarball does not match the integrity recorded in pnpm-lock.yaml. When a package is already locked with an integrity value, and the registry later serves different metadata and tarball content for the same package name and version, pnpm initially reports an integrity mismatch. However, plain pnpm install then performs a resolution repair, accepts the registry's new integrity, updates the lockfile, installs the new content, and exits successfully. This means the lockfile integrity check does not act as a hard stop by default. This vulnerability is fixed in 10.34.0 and 11.4.0.

pnpm is a package manager. Prior to 10.34.0 and 11.4.0, pnpm's tarball extraction worker skips integrity verification when the integrity field is absent from the lockfile resolution. If an attacker can both modify pnpm-lock.yaml to remove the integrity: field and cause the referenced registry URL to serve altered package content, pnpm install --frozen-lockfile can install the altered package without an integrity error. npm's npm ci enforces integrity by default; pnpm's behavior of silently skipping verification is a pnpm-specific fail-open gap. This vulnerability is fixed in 10.34.0 and 11.4.0.

pnpm is a package manager. Prior to 10.34.0 and 11.4.0, pnpm can send user-level unscoped npm authentication credentials to a registry chosen by a repository-local .npmrc file. In the reproduced case, the user's npm config contains a default registry and an unscoped _authToken. The repository does not provide a token-bearing auth line. It only sets registry= to a different registry URL. During normal pnpm metadata/install workflows, pnpm binds the user-origin unscoped credential to the repository-selected registry and sends it as an Authorization header. This vulnerability is fixed in 10.34.0 and 11.4.0.

pnpm is a package manager. Prior to 10.34.0 and 11.4.0, pnpm allows a transitive dependency alias from registry package metadata to contain path traversal segments. During install, pnpm later uses that alias as a filesystem path when linking dependency nodes. As a result, a registry package can cause `pnpm install --ignore-scripts` to replace paths in the current project with symlinks to attacker-controlled dependency package directories. This vulnerability is fixed in 10.34.0 and 11.4.0.

pnpm is a package manager. Prior to 10.34.0 and 11.4.0, pnpm's patch application pipeline (@pnpm/patch-package) performs no path validation on file paths extracted from .patch files. An attacker who contributes a malicious patch file via a pull request can write attacker-controlled content to or delete arbitrary files on the filesystem during pnpm install, as the user running the install. The diff --git header paths containing ../../ sequences traverse out of the package directory, and the traversal is difficult to catch in code review because patch file diff headers are opaque to most reviewers. This vulnerability is fixed in 10.34.0 and 11.4.0.

pnpm is a package manager. Prior to 10.34.0 and 11.4.0, pnpm passes the lockfile-controlled git resolution.commit value to git fetch without a -- separator or commit-format validation. For git dependencies fetched through the shallow-fetch path, a malicious lockfile can replace the expected 40-character commit hash with a Git option such as --upload-pack=<command>. For SSH and local transports, --upload-pack can execute the supplied command. HTTPS transports ignore --upload-pack, so the practical attack surface is primarily SSH or local git dependencies. This vulnerability is fixed in 10.34.0 and 11.4.0.

jq is a command-line JSON processor. Prior to 1.8.2,` jq --rawfile` can turn a handled oversized-string error into invalid-state reuse and a real heap out-of-bounds write in assertion-disabled builds. When jv_load_file(raw=1) reads an attacker-controlled file, it repeatedly appends file chunks to the same jv string accumulator. Once jv_string_append_buf() returns jv_invalid_with_msg("String too long"), the raw-file loop does not stop. If the file contains at least one more byte, the next loop iteration appends a new chunk to an object that is already invalid. With assertions enabled this aborts in jvp_string_ptr(). With assertions disabled, the invalid object is interpreted as a string object and ASan reports heap-buffer-overflow. This vulnerability is fixed in 1.8.2.

pnpm is a package manager. Prior to 10.33.4 and 11.0.7, a malicious codeload.github.com server can serve whatever tarball it wants and pnpm will install it regardless of the lockfile. The lockfile does not store the hash of the dependencies from https://codeload.github.com. This means that if this server was compromised or a person's machine configuration was compromised, pnpm would download and install these dependencies. This vulnerability is fixed in 10.33.4 and 11.0.7.

jq is a command-line JSON processor. Prior to 1.8.2, comparing two sufficiently deeply nested arrays with the == operator exhausts the C stack on jq's ordinary command-line surface, resulting in denial of service via stack exhaustion (uncontrolled recursion). The crash occurs in jq's recursive structural comparison code, with the recursion repeating through jvp_array_equal() and jv_equal() in src/jv.c when comparing deeply nested arrays; a nearby sort comparator path through jv_cmp() in src/jv_aux.c overflows the stack at a larger nesting depth from the same missing recursion guard. Anyone running jq comparisons on attacker-controlled deeply nested JSON values, or embedding jq in a context where untrusted data can reach the == comparison path, is affected. This vulnerability is fixed in 1.8.2.

X.509 trust-chain bypass (path-depth exhaustion) in the OpenSSL compatibility certificate verifier (wolfSSL_X509_verify_cert()). This affects only builds with --enable-opensslextra whose application calls X509_verify_cert() with caller-supplied untrusted intermediates; for those users it is critical, otherwise the library is unaffected. Native wolfSSL TLS/DTLS usage is not impacted. X509_verify_cert() returned success based only on the last verified link rather than on reaching a trust anchor: when the supplied chain is deeper than the verifier's maximum path depth (default 100), path building runs out of depth while still walking untrusted intermediates and the chain is accepted even though it never reaches a configured trust anchor, allowing acceptance of an attacker-controlled certificate. The default TLS handshake (WOLFSSL_VERIFY_PEER) is not affected; only applications doing manual or deferred verification through this API are.

A flaw was found in Keycloak Policy Enforcer. This vulnerability allows any authenticated user to bypass all authorization policies, including role, scope, and User-Managed Access (UMA) permission checks. By including the configured access-denied page path within a request URL, either as a path segment or a query parameter, an attacker can gain unauthorized access to protected resources.

A flaw was found in org.keycloak.authorization. An authenticated user with a granted User-Managed Access (UMA) permission ticket for one resource can exploit this by using a specific permission request prefix to bypass per-resource access control. This allows the user to gain unauthorized access to all resources of that type within the same resource server, even if they do not have a ticket for those specific resources. This vulnerability requires the resource server to be configured in PERMISSIVE policy enforcement mode and affects typed resources with ownerManagedAccess enabled, where no explicit policy protects the resource type. The primary consequence is unauthorized information disclosure or modification of resources.

A flaw was found in Keycloak's client registration service. A remote attacker, possessing a previously issued Registration Access Token (RAT), could exploit this vulnerability to re-enable a client that an administrator had explicitly disabled. This bypasses security controls, allowing the attacker to reset the client's secret and potentially regain privileged API access. The primary impact includes unauthorized information disclosure and potential integrity compromise.

A flaw was found in Keycloak. A missing authorization check in the GroupResource.addChild() endpoint within the Admin REST API allows an authenticated user with limited administrative privileges to reparent any existing group. When Fine-Grained Admin Permissions v2 (FGAPv2) is enabled, an attacker with management rights over a single low-privilege group can reparent a highly privileged group (such as one possessing the realm-admin role) under their managed group. Because group permissions follow a hierarchical structure, this action unauthorizedly grants the attacker management and password-reset capabilities over the members of the targeted privileged group. An attacker can exploit this to reset an administrator's password, compromise the account, and achieve a full realm takeover, leading to a complete compromise of confidentiality, integrity, and availability.

A flaw was found in Keycloak. A remote attacker with administrative privileges, specifically those with `manage-client` permission or access to client registration endpoints, could bypass client Uniform Resource Identifier (URI) validation. This is achieved by registering a malicious client with a specially crafted redirect URI using a case-insensitive `javascript:` or `data:` scheme. This Cross-Site Scripting (XSS) vulnerability allows for arbitrary code execution in the Keycloak origin when a victim clicks the crafted link, such as in the logout flow or the Admin Console.

A flaw was found in Keycloak. A realm administrator with the "manage-realm" role can exploit this vulnerability by submitting an arbitrary filesystem path as a keystore parameter when creating a key provider component. This allows the administrator to probe arbitrary filesystem paths, determining which files exist and are readable by the Keycloak process. This information disclosure could be used to identify high-value targets for follow-on attacks.

socat versions 1.8.0.0 through 1.8.1.1 contain a heap-based buffer overflow vulnerability that allows a malicious SOCKS5 proxy server to overwrite adjacent heap memory by exploiting a sign-extension flaw in the DOMAINNAME reply parser. During connection setup, the domain name length byte is read through a signed char field causing a negative bytes_to_read value that is implicitly converted to size_t, resulting in an unbounded heap write into the 262-byte reply buffer with attacker-controlled size and content.

Halo is an open source website building tool. Prior to 2.24.3, a path traversal vulnerability in the backup download endpoint allows authenticated administrators to read arbitrary files from the server filesystem. The backup download endpoint (GET /apis/console.api.migration.halo.run/v1alpha1/backups/{name}/files/{filename}) in MigrationServiceImpl.download() resolves the backup filename via Path.resolve() without validating that the resolved path stays within the designated backups directory. Also, the Backup creation endpoint (POST /apis/migration.halo.run/v1alpha1/backups) does not sanitize the status fields during creation This vulnerability is fixed in 2.24.3.

ToolJet is the open-source foundation am AI-native platform for building and deploying internal tools, workflows and AI agents. Prior to 3.20.178-lts, any authenticated user with builder role (free tier) can overwrite a globally-shared marketplace plugin with arbitrary JavaScript that executes server-side with full Node.js access (require, process). The malicious code runs whenever any user on the instance triggers a query using that plugin — achieving both RCE and supply-chain compromise of the entire ToolJet deployment. This vulnerability is fixed in 3.20.178-lts.

ToolJet is the open-source foundation am AI-native platform for building and deploying internal tools, workflows and AI agents. Prior to 3.20.178-lts, there's an SSRF in the RestAPI data source component. The RestAPI data source executes HTTP requests server-side, and its private IP filter only checks the hostname string — not the resolved IP. DNS names like 169.254.169.254.nip.io resolve to the Azure IMDS link-local address and bypass the filter entirely. This allows any authenticated user (free tier) to steal Azure managed identity tokens for the AKS production cluster. This vulnerability is fixed in 3.20.178-lts.

ToolJet is the open-source foundation am AI-native platform for building and deploying internal tools, workflows and AI agents. Prior to 3.20.1780-lts, the authenticated endpoint POST /api/data-sources/decrypt returns the decrypted plaintext for any credential whose credential_id is supplied in the request body. Unlike every neighbouring data-source route, this handler is not protected by ValidateDataSourceGuard, does not receive the calling @User(), and the underlying CredentialsService.getValue() looks the credential up by id only, with no organization scoping. As a result, any authenticated user of any organization can decrypt the data-source secrets of any other organization by supplying that organization's credential_id — a cross-tenant confidentiality breach. This vulnerability is fixed in 3.20.1780-lts.

Trivy is a security scanner. Prior to 0.71.1, when Trivy downloads an OCI artifact, it uses the org.opencontainers.image.title annotation from the artifact manifest as the destination filename without validation. An attacker who can make Trivy fetch an attacker-controlled artifact can supply a crafted annotation that resolves to a path outside the intended destination, causing Trivy to write the layer content to an arbitrary location on the host filesystem. This vulnerability is fixed in 0.71.1.

Outline is a service that allows for collaborative documentation. Prior to 1.8.0, the AuthenticationHelper.canAccess function uses ctx.originalUrl to verify if an API key or OAuth token has the required scopes for a request. It extracts the resource by splitting the URL by / and taking the last segment. However, it fails to strip the URL fragment (#). Because Koa's router uses ctx.path (which strips the fragment) for routing, an attacker can append a fragment containing a permitted path (e.g., #foo/api/documents.info) to a restricted endpoint (e.g., /api/documents.create). The router will route the request to the restricted endpoint, but canAccess will evaluate the permitted path in the fragment, bypassing the API key scope restrictions and allowing privilege escalation. This vulnerability is fixed in 1.8.0.

LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, the POST /api/auth/2fa/backup/regenerate endpoint regenerates all 2FA backup codes without requiring any TOTP token or existing backup code verification. An attacker with a stolen session token can silently replace a victim's backup codes and use them to bypass 2FA login or disable 2FA entirely. This vulnerability is fixed in 0.8.4-rc1.

LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, the fix for CVE-2025-7105 added forkIpLimiter and forkUserLimiter rate limiters to POST /api/convos/fork to prevent rapid-fire conversation duplication. However, the POST /api/convos/duplicate endpoint — which is in the same file and performs the exact same expensive database operations — was not given any rate limiter. An authenticated user can bypass the CVE-2025-7105 fix by using /duplicate instead of /fork to exhaust server resources. This vulnerability is fixed in 0.8.4-rc1.

LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, LibreChat allows users to configure custom OpenAI-compatible API endpoints by setting a baseURL. This URL is used to construct HTTP requests without any SSRF validation — no private IP check, no scheme restriction, no DNS pinning. An authenticated user can set baseURL to internal network addresses. This vulnerability is fixed in 0.8.4-rc1.

LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.5, LibreChat's MCP OAuth implementation does not validate that the resource parameter from OAuth Protected Resource metadata (RFC 9728) matches the configured MCP server URL, allowing a malicious MCP server to steal access tokens intended for a legitimate server. This vulnerability is fixed in 0.8.5.

LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, the DELETE /api/messages/:conversationId/:messageId endpoint allows any authenticated user to delete any other user's messages. The validateMessageReq middleware only validates that the conversationId belongs to the requesting user, but the handler calls deleteMessages({ messageId }) using only the messageId as the MongoDB filter — without adding a user constraint. An attacker provides their own valid conversationId (to pass validation) and the victim's messageId (to target deletion), resulting in permanent, irrecoverable message deletion. This vulnerability is fixed in 0.8.4-rc1.

LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, the POST /api/files/images endpoint allows any authenticated user to upload files into any agent's tool_resources (e.g., context, execute_code) without verifying ownership or EDIT permission on the target agent. A permission check was added to the POST /api/files route in a previous patch, but the image upload route was never updated with the same check. An attacker can simply use the image endpoint instead of the file endpoint to bypass the authorization entirely. This vulnerability is fixed in 0.8.4-rc1.

LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, there is a vulnerability in LibreChat's markdown artifact preview pipeline. The marked library v15.0.12 does not HTML-escape double-quote characters in image alt text when a custom renderer falls through to the default renderer. LibreChat's generateMarkdownHtml function (in client/src/utils/markdown.ts) installs a custom image renderer that returns false for URLs passing the isSafeUrl allowlist check, which causes marked to fall back to its built-in renderer. That built-in renderer inserts the raw alt text into the alt="..." attribute without escaping double-quote characters. An attacker can craft an alt text such as " onload="payload to break out of the attribute and inject an arbitrary event handler. The resulting HTML is then assigned to document.getElementById('content').innerHTML inside the Sandpack preview iframe, causing the payload to execute in the victim's browser. This vulnerability is fixed in 0.8.4-rc1.

LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, the fix for CVE-2024-11171 (commit bb58a2d0) added limits: { fileSize } to createMulterInstance() in the file upload routes. However, the POST /api/convos/import endpoint uses a separate multer instance that was never updated with the same limits configuration. Combined with the application-level size check being disabled by default (the CONVERSATION_IMPORT_MAX_FILE_SIZE_BYTES env var is commented out in .env.example), an authenticated user can upload arbitrarily large files to exhaust server disk space and memory. This vulnerability is fixed in 0.8.4-rc1.

HTMLy CMS through 3.1.1 contains a path traversal vulnerability that allows low-privileged authenticated attackers to relocate arbitrary files by supplying directory traversal sequences in the oldfile parameter at the admin autosave endpoint. Attackers can pass unsanitized traversal sequences directly to file_exists() and rename() functions in admin.php without canonicalization or directory boundary enforcement to cause unintended relocation of any file writable by the web server process to an attacker-specified draft location.

Zephyr's IPv6 network stack can be prevented from receiving or processing future incoming packets by sending a small number of maliciously fragmented IPv6 packets. When such a packet is handled by the fragment-header processing path, the associated RX network packet buffer (allocated from a memory slab) is not released back to the pool. Repeating the malicious packet exhausts all RX buffer slots, after which the device can no longer obtain RX buffers and stops receiving traffic, resulting in a denial of service.

Permissions where checked incorrectly during room creation, allowing attackers to create rooms of types they shouldn't be allowed to create.