The `BetaLocalFilesystemMemoryTool` in the Anthropic TypeScript SDK created memory files and directories using the Node.js default modes (`0o666` for files, `0o777` for directories), leaving them world-readable on systems with a standard umask and world-writable in environments with a permissive umask such as many Docker base images. A local attacker on a shared host could read persisted agent sta

### Summary Versions of `i18next-http-middleware` prior to 3.9.3 pass the user-controlled `lng` and `ns` values from `getResourcesHandler` directly into `i18next.services.backendConnector.load(languages, namespaces, …)` without any sanitisation. Depending on which backend is configured, the unvalidated path segments enable one of two attacks: - **Filesystem path traversal** when the middleware i

### Summary A critical Denial of Service (DoS) vulnerability exists in `marked@18.0.0`. By providing a specific 3-byte input sequence a tab, a vertical tab, and a newline (`\x09\x0b\n`)—an unauthenticated attacker can trigger an infinite recursion loop during parsing. This leads to unbounded memory allocation, causing the host Node.js application to crash via Memory Exhaustion (OOM). ### Details

## Impact OpenClaw deployments before `2026.4.15` could embed host-local audio files into webchat responses without applying the local media root containment check used by other media-serving paths. If an attacker could influence an agent or tool-produced `ReplyPayload.mediaUrl`, the webchat audio embedding helper could resolve an absolute local path or `file:` URL, read an audio-like file under

## Impact OpenClaw deployments before `2026.4.21` could treat a non-owner sender as authorized for owner-enforced slash commands when all of the following were true: - a channel plugin declared `commands.enforceOwnerForCommands: true`; - the channel accepted wildcard inbound senders with `allowFrom: ["*"]`; - no explicit `commands.ownerAllowFrom` was configured. In that state, `src/auto-reply/c

## Impact An authenticated user with permission to create or modify workflows could achieve global prototype pollution via the XML Node leading to RCE when combined with other nodes exploiting the prototype pollution. ## Patches The issue has been fixed in n8n versions 1.123.32, 2.17.4, and 2.18.1. Users should upgrade to one of these versions or later to remediate the vulnerability. ## Workarou

## Impact A flaw in the `xml2js` library used to parse XML request bodies in n8n's webhook handler allowed prototype pollution via a crafted XML payload. An authenticated user with permission to create or modify workflows could exploit this to pollute the JavaScript object prototype and, by chaining the pollution with the Git node's SSH operations, achieve remote code execution on the n8n host. #

## Impact An unauthenticated attacker could register a malicious MCP OAuth client with a crafted `client_name`. If a victim user authorized the OAuth consent dialog and a second user subsequently revoked that access, a toast notification would render the injected script. Clicking the link would execute arbitrary JavaScript in the victim's authenticated n8n browser session, enabling credential and

## Impact The `dynamic-node-parameters` endpoints did not verify whether the authenticated caller was authorized to use a supplied credential reference. An authenticated user with access to a shared workflow could supply a foreign credential ID in the request body, causing the backend to decrypt and use that credential in a helper execution path where the caller also controls the destination URL.

## Impact An authenticated user with permission to create or modify workflows containing a Python Code Node could escape the sandbox and achieve arbitrary code execution on the task runner container. - This issue only affects instances where the Python Task Runner is enabled. ## Patches The issue has been fixed in n8n versions 1.123.32, 2.17.4, and 2.18.1. Users should upgrade to one of these ve

## Impact An authenticated user with a valid API key scoped to `variable:list` could read variables from projects they are not a member of by supplying an arbitrary `projectId` query parameter to the public API variables endpoint. The handler queried the variables repository directly without enforcing project membership checks, bypassing the authorization-aware service layer used by the internal e

## Impact The MCP OAuth client registration endpoint accepted unauthenticated requests and stored client data without adequate resource controls. An unauthenticated remote attacker could exhaust server memory resources by sending large registration payloads, rendering the n8n instance unavailable. The MCP enable/disable toggle gates MCP access but did not restrict client registrations, meaning the

## Impact The `/chat` WebSocket endpoint used by the Chat Trigger node's Hosted Chat feature did not verify that an incoming connection was authorized to interact with the target execution. An unauthenticated remote attacker who could identify a valid execution ID for a workflow in a waiting state could attach to that execution, receive the pending prompt intended for the legitimate user, and subm

## Impact A flaw in the SeaTable node's `row:search` and `row:get` operations allowed user-controlled input to be concatenated directly into SQL query strings without escaping or parameterization. In workflows where external user input is passed via expressions into the SeaTable node's search or row retrieval parameters, an attacker could manipulate the constructed query to retrieve unintended row

## Impact The `/mcp-oauth/register` endpoint accepted OAuth client registrations without authentication, allowing arbitrary `redirect_uri` values to be registered. When a user denies the MCP OAuth consent dialog, the `handleDeny` handler redirects the user to the registered `redirect_uri` without validation, enabling an open redirect to an attacker-controlled URL. An attacker can craft a phishing

## Impact A flaw in the Oracle Database node's select operation allowed user-controlled input passed into the `Limit` field via expressions to be interpolated directly into the SQL query without sanitization or parameterization. In workflows where external input is passed into the `Limit` field (e.g., from a webhook), an attacker could inject arbitrary SQL and exfiltrate data from the connected Or

## Impact The fix for [GHSA-f3f2-mcxc-pwjx](https://github.com/advisories/GHSA-f3f2-mcxc-pwjx) did not cover the Snowflake node or the legacy MySQL v1 node. Both nodes construct SQL queries by directly interpolating user-controlled table names, column names, and update keys into query strings without identifier escaping, enabling SQL injection against the connected database. Exploitation requires