re2 provides Node.js bindings for Google's RE2 regular expression engine. Prior to 1.25.2, re2 validates lastIndex against the UTF-8 byte length of a subject but uses it as a UTF-16 code-unit offset in exec, test, match, replace, and split, allowing an attacker-influenced lastIndex on a non-ASCII subject to trigger an out-of-bounds heap read and an uncatchable process crash, with limited heap information disclosure in some cases. This issue is fixed in 1.25.2.

WACRM is a self-hostable CRM template for WhatsApp. In 0.7.0 and earlier, the automation send_webhook action in src/lib/automations/engine.ts and its validation in src/lib/automations/validate.ts allowed an authenticated user with automation privileges to submit an arbitrary webhook URL that the server fetched without the existing isDeliverableUrl SSRF guard in src/lib/webhooks/ssrf.ts, allowing requests to private, loopback, link-local, or cloud metadata addresses such as the cloud metadata endpoint at 169.254.169.254. This vulnerability is fixed with commit 23838a9959550e975d732ae08a44a3a2f0cc084b.

OpenProject is open-source, web-based project management software. Prior to 17.6.0, PATCH /api/v3/work_packages/{id} accepted _links.fileLinks and allowed authenticated users with edit_work_packages but without manage_file_links to resolve Storages::FileLink records by raw id, detach or hard-delete existing FileLinks, and re-parent FileLinks from other projects to an attacker-controlled work package, exposing origin filename, origin id, and MIME type metadata. This issue is fixed in 17.6.0.

Juggle through 1.6.0 contains a remote code execution vulnerability that allows unauthenticated remote attackers to execute arbitrary OS commands by connecting to the exposed H2 database web console using default shipped credentials. Attackers can access the unprotected /h2-console endpoint, authenticate with default credentials, and leverage the H2 CREATE ALIAS Runtime.exec() technique to execute arbitrary commands, resulting in root-level code execution when running the stock Docker image.

Wolf CMS through 0.8.3.1 contains a remote code execution vulnerability in FileManagerController that allows authenticated attackers to create arbitrary PHP files by exploiting missing file extension validation in the create_file() and save() functions. Attackers with the file_manager_mkfile capability can write malicious PHP content into the web-accessible FILES_DIR directory and trigger execution by requesting the file over HTTP.

Capsule is a multi-tenancy and policy-based framework for Kubernetes. From 0.13.0 until 0.13.8, after the incomplete CVE-2026-22872 fix, TenantResource RawItems and Generators in internal/controllers/resources/collect.go, including handleRawItem and handleGeneratorItem, did not apply the ResourceReference.LoadResources and IsNamespacedGVK cluster-scoped resource rejection guard used by NamespacedItems, allowing a Tenant Owner to create cluster-scoped resources such as ClusterRole or ValidatingWebhookConfiguration through the cluster-admin controller client. This issue is fixed in version 0.13.8.

CleverTap Web SDK v1.15.1 is vulnerable to Cross Site Scripting (XSS). The application does not sanitize untrusted data received via window.postMessage before injecting it into the page DOM. An attacker can craft a malicious message that, when processed by renderCustomHtml, results in execution of arbitrary JavaScript in the context of the hosting site.

Leantime 3.6.2 contains a cross-site request forgery vulnerability that allows unauthenticated attackers to perform state-changing actions on behalf of authenticated users by excluding the Laravel VerifyCsrfToken middleware from the global middleware stack in app/Http/Kernel.php. Attackers can craft malicious pages delivered via phishing emails or malicious websites to trigger unauthorized POST, PUT, and DELETE requests that create or delete projects, modify settings, and change permissions as any authenticated user.

Banks generates meaningful LLM prompts using a simple template language. In versions prior to 2.4.3, banks parses Tool JSON objects from the rendered body of {% completion %} blocks and later resolves their import_path field through importlib.import_module(...) + getattr(...) to obtain the callable that handles a tool call. There is no allowlist or sanitization on import_path, so any importable Python attribute (e.g. os.system, subprocess.getoutput) can be selected. When the LLM emits a tool_calls entry whose function.name matches the attacker-supplied tool name, the resolved callable is invoked with kwargs decoded from tool_call.function.arguments, yielding arbitrary code execution in the banks-hosting process. This is distinct from GHSA-gphh-9q3h-jgpp / CVE-2026-44209. That advisory was fixed in 2.4.2 by switching src/banks/env.py from Environment to SandboxedEnvironment. The fix does not touch src/banks/extensions/completion.py, and the unsafe import + getattr chain still executes on 2.4.2. The malicious Tool JSON is plain text in the rendered template body — it requires no Jinja attribute access, so the sandbox is irrelevant. This issue has been fixed in version 2.4.3.

An API session‑management flaw in products with the MikroTik RouterOS API enabled are vulnerable to a Insufficient Session Expiration vulnerability. This could allow active sessions to retain their previous permission set after inactivity timeouts or user‑group changes. As a result, an authenticated user whose permissions have been reduced may continue accessing information.

IBM Engineering Requirements Management DOORS and DOORS Web Access 9.7.2.1 through 9.7.2.11, and 9.6.1.1 through 9.6.1.13 is vulnerable to cross-site scripting. This vulnerability allows an unauthenticated attacker to embed arbitrary JavaScript code in the Web UI thus altering the intended functionality potentially leading to credentials disclosure within a trusted session.

Leantime 3.6.2 contains an open redirect vulnerability in the Login controller that allows unauthenticated attackers to redirect authenticated users to arbitrary external sites by manipulating the redirectUrl POST parameter. Attackers can craft a malicious login URL with a tampered redirectUrl value that bypasses FILTER_SANITIZE_URL validation to redirect victims to attacker-controlled sites for phishing or credential theft.

A security flaw combining LDAP filter injection and improper authorization checks was found in Samba Active Directory Domain Controller (AD DC). When processing LDAP Compare requests, Samba fails to properly validate user-supplied attribute names and executes the resulting internal database search in a trusted context, bypassing normal Access Control List (ACL) enforcement. An authenticated low-privilege domain user can exploit these flaws to disclose confidential Active Directory attributes that would normally be inaccessible. The disclosed information may be leveraged to derive sensitive authentication material, potentially leading to privilege escalation and complete domain compromise. For example: In deployments configured with Group Managed Service Accounts (gMSAs), an attacker can extract the "msKds-RootKeyData" attribute and derive gMSA passwords offline, potentially leading to complete domain compromise if privileged gMSAs are present.

An out-of-bounds read flaw was found in Samba's Kerberos Key Distribution Center's (KDC) password change (kpasswd) service. When processing malformed ASN.1-encoded Kerberos password change request, Samba server miscalculates the structure size and attempts to read up to six bytes beyond the end of the allocated buffer. While this out-of-bounds read typically results in a harmless decryption failure, if the read hits unmapped memory, it causes the KDC process to crash. An authenticated attacker can send a specially crafted kpasswd request containing malformed ASN.1 data to trigger the out-of-bounds read, which may cause the KDC process to terminate, resulting in a denial of service.

Kanboard 1.2.52 and prior contains a server-side request forgery vulnerability that allows authenticated users to bypass SSRF protections by supplying hexadecimal IP address notation in user-controlled URLs. Attackers can submit hexadecimal-encoded internal IP addresses through the web link creation feature, causing cURL to resolve and connect to internal network resources such as cloud instance metadata services, localhost services, and RFC1918 addresses while the isPrivateURL() filter in app/Core/Http/Client.php incorrectly treats the input as safe due to FILTER_VALIDATE_IP rejecting non-dotted-decimal notation.

SGLang contains a model weight exfiltration vulnerability when no API keys are configured, as SGLang will expose two endpoints that allow a remote attacker to trigger distributed weight broadcasting using NCCL and then triggering data transfer, attackers can exfiltrate all model weights.

lmsys

SGLang contains a RCE vulnerability when attempting to load model weights from a HuggingFace repository, specifically within the /update_weights_from_disk, where torch.load(..., weights_only=False) fallback enables pickle deserialization of .bin files.

lmsys

IBM Langflow OSS 1.0.0 through 1.10.1 can allow an attacker to access another user's private vector documents by creating their own flow with matching Chroma persist_directory and collection_name values. The attacker receives exact victim content in their workflow output despite having no authorization to read the victim's flow. Additionally, the attacker can pollute the victim's collection by inserting their own documents into the shared namespace.

langflow

Calico's Application Layer Policy (disabled by default), which enforces HTTP rules through Dikastes, fails to perform URL path normalization. As a result, HTTP requests using path-traversal segments, encoded slashes, or repeated slashes are not correctly evaluated by Prefix path rules. Dikastes authorizes the request under the permitted prefix while the downstream workload or a fronting proxy normalizes the path and serves the restricted endpoint. An attacker with network access and no special RBAC can potentially reach HTTP endpoints the policy was intended to restrict.

MaxKey through 4.1.12, fixed in commit ddbb72f, contains an insufficient redirect URI validation vulnerability in DefaultRedirectResolver.hostMatches() that allows remote attackers to hijack OAuth 2.0 authorization codes by supplying a crafted redirect_uri whose hostname suffix matches a registered URI without proper dot-boundary anchoring. Attackers who control a domain ending with the registered redirect URI hostname can social-engineer victims into clicking a crafted authorization URL, causing the authorization code to be issued to the attacker-controlled URI and exchanged for an access token granting access to the victim's identity.

Improper Isolation or Compartmentalization vulnerability in malach-it boruta (Elixir.Boruta.Openid module) allows attackers to register OpenID Connect clients with administrative privileges through the dynamic client registration entry point. Boruta.Openid.register_client/3 forwards caller-supplied registration parameters to the administrative client creation path without a public/admin field-level allowlist, so an unauthenticated registrant can set security-sensitive attributes including supported grant types, authorized scopes, PKCE enforcement, public refresh and revocation behavior, token lifetimes, and signing settings. The library does not distinguish between metadata a public registrant is allowed to set and administrative controls that should require operator approval. This vulnerability is associated with program files lib/boruta/openid.ex and program routines 'Elixir.Boruta.Openid':register_client/3, 'Elixir.Boruta.Openid':parse_registration_params/2. This issue affects boruta from 2.3.0 before 2.3.7.