Threat briefs

The `repomix` CLI tool is vulnerable to command injection (CVE-2026-49987) via unsanitized user input in the `--remote-branch` argument, allowing attackers to inject arbitrary `git` command-line options like `--upload-pack` and achieve remote code execution with the privileges of the running user, potentially leading to CI/CD pipeline compromise.

A high-severity vulnerability (CVE-2026-48815) in the `npm/sigstore` library (versions <= 4.1.0) causes the `certificateOIDs` verification constraint to be silently ignored, allowing applications to accept unauthorized certificates that should have been rejected based on extension policy, which could lead to supply chain attacks by trusting malicious artifacts.

A critical vulnerability (versions prior to 3.1.0) in SurrealDB's HTTP /rpc endpoint allowed unauthenticated attackers to enumerate session UUIDs via the `sessions` method, enabling full session hijack of any attached and authenticated session due to a lack of ownership checks, leading to unauthorized data manipulation and privilege escalation.

A critical authentication bypass vulnerability exists in Centrifugo v6's dynamic JWKS endpoint feature, allowing an attacker to bypass JWT authentication for one tenant by leveraging a valid token from another tenant due to incorrect JWKS key caching indexed only by the `kid`.

A vulnerability in Fleet's agent-side deployer (CVE-2026-44938) allows an attacker with `git push` access to a Fleet-monitored repository to overwrite Pod Security Standards (PSS) enforcement labels on target Kubernetes namespaces, bypassing admission controls and enabling the deployment of otherwise prohibited workloads.

An unauthenticated regex injection vulnerability exists in Rancher Fleet's webhook endpoint when it's configured without a secret, allowing attackers to forge webhook requests using unsanitized repository URL components, which leads to continuous repository re-cloning, causing network and resource exhaustion (Denial of Service) on the management cluster, and potentially service downgrades if the attacker has read access to the target Git repository.

An attacker can craft a malicious OCI artifact with a tarball layer containing a hardlink entry that uses a relative path for its target, which, when extracted by `oras-go` (<= 2.6.1) or the `oras` CLI, allows the hardlink to resolve against the process's current working directory (CWD) instead of the intended extraction base, leading to arbitrary file read or modification in the victim's CWD via an inode-sharing vulnerability.

A vulnerability (CVE-2026-50138) in `goshs` versions up to `v2.0.9` allows an authenticated attacker to bypass intended access restriction flags like `--read-only`, `--upload-only`, and `--no-delete` when the WebDAV listener is enabled, leading to unauthorized file creation, modification, deletion, and content exfiltration on the server, compromising data integrity and confidentiality.

An attacker can exploit a Server-Side Request Forgery (SSRF) vulnerability in `@apify/actors-mcp-server` version `0.10.7` by crafting a malicious Actor definition to inject an arbitrary authority into a URL, causing the MCP client to exfiltrate the victim's Apify API token to the attacker's server, granting full access to their Apify account.

A high-severity vulnerability in OpenClaw's exec approval feature (versions prior to 2026.5.18) allows an authenticated attacker to bypass approval integrity by crafting a long command that appears benign in the truncated UI but contains a malicious suffix, leading to unauthorized command execution.

A high-severity vulnerability, CVE-2026-53810, in OpenClaw's marketplace runtime extension metadata allows an attacker to craft a malicious package that, when installed by a trusted operator, redirects runtime loading to hidden, unscanned code, potentially leading to unauthorized code execution and bypassing security checks.

A vulnerability (GHSA-rggc-m335-3wvj) in OpenClaw's trusted-proxy deployments allows a local attacker on the same host to forge identity headers, bypassing intended security controls and potentially leading to unauthorized access or privilege escalation if the affected feature is enabled and reachable.

An authentication bypass vulnerability (CVE-2026-53817) in OpenClaw's Control UI pairing mechanism allows an attacker with existing network/authentication foothold in LAN/shared-token deployments to spoof locality information, leading to the acquisition of a durable admin-capable device token that grants persistent administrative access, even after shared gateway tokens are rotated.

A vulnerability in OpenClaw's `system.run` safe-bin feature on POSIX nodes could allow a lower-privilege operator flow to read local files not intended by policy, as shell expansion can alter the interpretation of an approved command, causing a seemingly safe argument to expand into additional shell words and become a file operand, potentially exposing OpenClaw configuration data or other node-local information.

A vulnerability in OpenClaw allows an attacker with `operator.write` privileges to bypass intended administrative command scope gates by delivering a scoped Gateway `chat.send` request through an inherited external route, leading to unauthorized execution of critical administrative commands.

A high-severity vulnerability (CVE-2026-53811) in OpenClaw's Matrix `allowFrom` feature allows threat actors to exploit mutable display names to match policy entries, potentially granting unauthorized agent access intended for another Matrix identity.

A high-severity vulnerability (GHSA-j472-gf56-x589) in OpenClaw allows an attacker to bypass allowlist checks for PowerShell encoded commands by using abbreviated encoded-command flags, leading to unauthorized code execution on the underlying Windows system if a vulnerable feature is enabled and reachable.

A path traversal vulnerability (CVE-2026-50181) exists in Langroid's `ReadFileTool` and `WriteFileTool` components (versions <= 0.63.0), allowing an attacker to read or write arbitrary files outside the configured `curr_dir` via crafted `file_path` arguments, potentially leading to sensitive information disclosure or unauthorized file modification in applications exposing these tools to user or LLM input.

A high-severity vulnerability (GHSA-c29c-2q9c-pc86) in OpenClaw's handling of Slack's `allowFrom` feature could allow an attacker to gain unintended agent access by manipulating their Slack display name metadata to match a policy entry, especially in configurations where the affected feature is enabled and reachable.

A vulnerability in OpenClaw's trusted-proxy Control UI mode allows an unpaired or restricted trusted-proxy client to gain temporary `operator.admin` authority by declaring elevated WebSocket scopes before proper server-side authorization, enabling the execution of admin-gated Gateway RPCs until the connection is closed or revalidated.

A high-severity vulnerability (affecting OpenClaw versions prior to 2026.5.4) in the bundled device-pair plugin allowed authorized non-owner chat senders to issue device-pairing bootstrap codes, enabling them to enroll devices with operator/node capabilities and gain persistent unauthorized access within the OpenClaw environment.

A high-severity vulnerability (CVE-2026-53819) in OpenClaw versions prior to 2026.5.27 allows a malicious `.env` file within a repository to override the Homebrew executable selection during skill installation flows, potentially leading to arbitrary code execution on trusted operator systems running macOS or Linux.

A high-severity vulnerability, CVE-2026-53813, in npm/openclaw versions <= 2026.4.24 allows fake package roots to influence memory-core artifact loading, potentially leading to the selection and execution of unintended local artifacts based on attacker-controlled or lower-trust input reaching the affected path.

A high-severity vulnerability, CVE-2026-53806, in npm/openclaw versions up to 2026.5.7, allows attackers to bypass 'exec revalidation' controls by confusing the application with combined POSIX shell options, leading to unauthorized inline shell content execution and potential remote code execution.

A vulnerability, CVE-2026-53816, in npm/openclaw versions prior to 2026.5.18, allows a malicious or compromised paired node to forge 'exec' lifecycle events and send them to the gateway, which, due to a missing provenance check, accepts the attacker-supplied event data as legitimate execution results, leading to unauthorized capability exposure for the compromised node.

A high-severity vulnerability (GHSA-w5ww-7chg-mxcq) in OpenClaw allows an unauthorized Telegram user to bypass the `commands.allowFrom` sender check via interactive callbacks, leading to unauthorized command execution on the OpenClaw Gateway.

A vulnerability in OpenClaw allows an attacker to bypass trusted retry endpoint validation by crafting a URL with a hostname prefix that resembles a trusted host, which, if the feature is enabled and reachable by lower-trust input, could lead to sensitive authentication material being sent to an unintended external endpoint.

A high-severity stored cross-site scripting (XSS) vulnerability, CVE-2026-52854, exists in the MediaWiki Maps extension (versions prior to 12.1.3), allowing any authenticated user with edit permissions to inject malicious JavaScript into the `overlays` parameter of the `display_map` parser function, leading to arbitrary client-side code execution in a victim's browser.

A command injection vulnerability (CVE-2026-44454) in the Coder platform's `dotfiles` module allows arbitrary code execution in a user's workspace, exploitable via a one-click attack using the `mode=auto` feature on the Create Workspace page that automatically provisions a workspace with a malicious `param.dotfiles_uri` without user consent, leading to immediate arbitrary code execution and potential data compromise or lateral movement.

A low-privileged authenticated user can exploit CVE-2026-50279, an authorization bypass vulnerability in Craft CMS's `entries/save-entry` endpoint, to reassign an entry's authorship to another user without proper permissions, leading to corrupted audit trails and misleading content ownership.