Codefriar sf-mcp versions up to 1.3.2 contain an OS command injection vulnerability. This flaw arises because the escapeValue() function does not properly neutralize shell metacharacters such as backticks and $() substitution. As a result, crafted Salesforce CLI tool flag values can be executed as shell commands with the privileges of the MCP server user.
CVE-2026-117558 is an OS command injection vulnerability in the ai-sdlc plugin MCP server versions up to 0.31.2. It occurs because the repo argument is not properly validated and is interpolated into shell commands executed via execSync(), allowing attackers to execute arbitrary OS commands with the privileges of the MCP server user.
CVE-2026-116369 is an OS command injection vulnerability in the github-kanban-mcp-server by Sunwood-ai-labs, affecting versions 0.1 through 0.4.0. The flaw exists in the handleListIssues function in src/handlers/issue-handlers.ts, where manipulation of the state or labels arguments can lead to command injection. This vulnerability can be exploited remotely. Upgrading to version 0.4.1 resolves the issue.
FastMCP versions 3.0.0 through 3.4.9 and 4.0.0 through 4.1.1 contain an authorization bypass vulnerability where direct calls to transformed components skip the final authorization check. This allows low-privilege users to invoke protected tools, resources, templates, or prompts that should be denied. The issue arises because the server applies transforms before authorization but does not re-check the authorization on the transformed component in direct lookup paths. Fixes are available in FastMCP 3.4.10 and 4.1.2.
FastMCP versions 3.0.0–3.4.9 and 4.0.0–4.1.1 have an authorization vulnerability where direct calls to transformed components skip the final authorization check. This allows low-privilege users to invoke protected tools, resources, templates, or prompts that should be denied. The issue arises because the server applies transforms to components but does not reauthorize the transformed component before returning it on direct lookups. Fixes are available in FastMCP 3.4.10 and 4.1.2.
CVE-2026-116109 is an OS command injection vulnerability in continuedev Continue version 2.0.0. It affects the MCPManagerSingleton.setConnections function in the MCP Server Transport component. The vulnerability allows remote attackers to execute arbitrary OS commands. Public exploit code has been disclosed, but the vendor has not responded or issued a fix.
CVE-2026-87912 and CVE-2026-87913 refer to missing S3 bucket ownership verification vulnerabilities in the AWS Security Agent plugin for aws-agents-for-devsecops and MCP Server. The vulnerabilities relate to improper validation of S3 bucket ownership, which could potentially lead to security issues. No specific affected versions or detailed technical information are provided. No known exploits are reported in the wild. No patch or remediation information is available at this time.
CVE-2026-85787 is a vulnerability in the SQL validation component of Amazon awslabs postgres-mcp-server versions prior to 1.1.7. It allows an unauthenticated actor to modify data beyond the intended read-only scope by injecting crafted SQL through content submitted by an authenticated user. The issue is fixed in version 1.1.7. Mitigation includes upgrading to version 1.1.7 or later and running the MCP server with a minimal-privilege Postgres role to enforce database-level access controls.
An incomplete list of disallowed inputs in the Python code validator of the run_script tool in AWS MCP Server might allow an authenticated remote user to execute arbitrary code within their own isolated session environment via a crafted script.
AWS fixed this issue service-side. No customer action is required.
Uni-CLI versions prior to 0.225.2 exposed a legacy JSON-RPC-over-HTTP transport on localhost without validating browser Origin headers. This allowed malicious web pages to send CORS POST requests to the local /mcp endpoint and execute JSON-RPC commands against the local Uni-CLI server if the MCP HTTP transport was active. The issue was due to inconsistent enforcement of the browser-to-localhost boundary between two HTTP transports. This vulnerability has been fixed in version 0.225.2.
Agentfront mcp-from-openapi versions from 2.2.0 up to but not including 2.9.0 contain a server-side request forgery (SSRF) vulnerability. This flaw allows attackers to bypass the SSRF blocked hosts denylist by appending a trailing dot to hostnames, enabling unauthorized requests to operator-blocked hosts. The vulnerability is fixed in version 2.9.0.
CVE-2026-114612 is a critical improper privilege management vulnerability in BerriAI's LiteLLM proxy server versions from 1.87.0 up to certain patched releases. The flaw allows an authenticated internal user to forge administrator credentials by exploiting reuse of a cryptographic context for secrets and tokens. This enables escalation to administrative access and execution of arbitrary commands via the MCP stdio endpoint. The issue affects versions 1.87.0 through 1.100.4 and several subsequent minor versions, with fixes released in 1.100.4 and later.
The rmcp OAuth client versions from 2.0.0 up to but not including 3.7.0 have a server-side request forgery (SSRF) vulnerability. The client only inspected host strings to decide if it should fetch authorization server metadata, failing to resolve hostnames to their IP addresses. This allowed malicious servers to cause the client to send requests to internal IP addresses, potentially reaching unauthenticated internal HTTP services. The issue was fixed in version 3.7.0 by adding DNS resolution and address policy checks on all resolved addresses and redirects.
CVE-2026-114123 is a vulnerability in dynamics365ninja d365fo-mcp-server versions prior to 1.11.0 where missing authentication allows unauthenticated access to critical read-only API endpoints. The default Azure deployment sets the API key to an empty string, causing the server to accept requests without authentication. This exposes proprietary X++ source code snippets and metadata, including business logic and security role information. The issue is fixed in version 1.11.0.
An integer underflow vulnerability exists in the mcpack2pb component of Apache bRPC versions prior to 1.18.0. This flaw allows remote attackers to crash a bRPC server by sending a specially crafted request with a string field having a zero value_size. The vulnerability arises because the UnparsedValue::as_string() function subtracts one from the value_size without validation, causing an underflow and an uncaught exception that aborts the process. Default server configurations are not affected as the vulnerable protocol is registered client-only by default.
Apache bRPC versions prior to 1.18.0 contain an uncontrolled resource allocation vulnerability in the mcpack2pb component. This flaw allows remote attackers to cause the server to allocate excessive memory by sending a specially crafted mcpack request with an abnormally large array item count. The vulnerability arises because the mcpack2pb parser trusts the declared item count without validating it against the actual payload size, leading to potential process aborts due to memory exhaustion.
Uncontrolled recursion in the mcpack2pb component in Apache bRPC (version < 1.18.0) on all platforms allows remote attackers to make the server crash via sending deeply nested mcpack data.
Root Cause:
The bRPC mcpack2pb parser (used to serve nshead_mcpack / ubrpc / public_pbrpc requests) recursively parses nested objects and arrays without any depth limit. For a protobuf message that references itself (e.g. a repeated message field), the generated parser recurses once per nesting level of the incoming mcpack object. A deeply nested request makes the recursion grow unbounded, overflow the stack and crash the process.
Affected Scenarios:
Use bRPC server with a recursive-schema protobuf message to serve nshead_mcpack / ubrpc / public_pbrpc requests from untrusted network.
How to Fix:
(Choose one of the following options)
1. Upgrade bRPC to version 1.18.0, which fixes this issue.
2. Apply this patch: https://github.com/apache/brpc/pull/3499
Note:
No matter which option you choose, you should know that the fix enforces the documented parsing depth limit MAX_DEPTH (128) on the parse path, the same limit the serializer already applies. If your requests contain mcpack messages nested deeper than 128 levels, the request will be failed after applying the fix.
Trigger.dev cli-v3 from 3.0.0 before 4.5.4 contains a missing authentication vulnerability in the trigger dev --mcp server, which binds all network interfaces and exposes the /sse and /messages endpoints. LAN attackers or malicious web pages can send MCP tools/call requests to invoke trigger-task, cancel-run, list-runs, and get-run-logs using the developer's stored access token.