agpt CVE Vulnerabilities & Metrics

AutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows. The AutoGPT Platform's WebSocket API transmitted node execution updates to subscribers based on the graph_id+graph_version. Additionally, there was no check prohibiting users from subscribing with another user's graph_id+graph_version. As a result, node execution updates from one user's graph execution could be received by another user within the same instance. This vulnerability does not occur between different instances or between users and non-users of the platform. Single-user instances are not affected. In private instances with a user white-list, the impact is limited by the fact that all potential unintended recipients of these node execution updates must have been admitted by the administrator. This vulnerability is fixed in 0.6.1.

AutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows. Prior to 0.6.1, AutoGPT allows of leakage of cross-domain cookies and protected headers in requests redirect. AutoGPT uses a wrapper around the requests python library, located in autogpt_platform/backend/backend/util/request.py. In this wrapper, redirects are specifically NOT followed for the first request. If the wrapper is used with allow_redirects set to True (which is the default), any redirect is not followed by the initial request, but rather re-requested by the wrapper using the new location. However, there is a fundamental flaw in manually re-requesting the new location: it does not account for security-sensitive headers which should not be sent cross-origin, such as the Authorization and Proxy-Authorization header, and cookies. For example in autogpt_platform/backend/backend/blocks/github/_api.py, an Authorization header is set when retrieving data from the GitHub API. However, if GitHub suffers from an open redirect vulnerability (such as the made-up example of https://api.github.com/repos/{owner}/{repo}/issues/comments/{comment_id}/../../../../../redirect/?url=https://joshua.hu/), and the script can be coerced into visiting it with the Authorization header, the GitHub credentials in the Authorization header will be leaked. This allows leaking auth headers and private cookies. This vulnerability is fixed in 0.6.1.

AutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows. Prior to 0.6.1, AutoGPT allows SSRF due to DNS Rebinding in requests wrapper. AutoGPT is built with a wrapper around Python's requests library, hardening the application against SSRF. The code for this wrapper can be found in autogpt_platform/backend/backend/util/request.py. The requested hostname of a URL which is being requested is validated, ensuring that it does not resolve to any local ipv4 or ipv6 addresses. However, this check is not sufficient, as a DNS server may initially respond with a non-blocked address, with a TTL of 0. This means that the initial resolution would appear as a non-blocked address. In this case, validate_url() will return the url as successful. After validate_url() has successfully returned the url, the url is then passed to the real request() function. When the real request() function is called with the validated url, request() will once again resolve the address of the hostname, because the record will not have been cached (due to TTL 0). This resolution may be in the "invalid range". This type of attack is called a "DNS Rebinding Attack". This vulnerability is fixed in 0.6.1.

AutoGPT versions 0.3.4 and earlier are vulnerable to a Server-Side Template Injection (SSTI) that could lead to Remote Code Execution (RCE). The vulnerability arises from the improper handling of user-supplied format strings in the `AgentOutputBlock` implementation, where malicious input is passed to the Jinja2 templating engine without adequate security measures. Attackers can exploit this flaw to execute arbitrary commands on the host system. The issue is fixed in version 0.4.0.

A command injection vulnerability exists in the workflow-checker.yml workflow of significant-gravitas/autogpt. The untrusted user input `github.head.ref` is used insecurely, allowing an attacker to inject arbitrary commands. This vulnerability affects versions up to and including the latest version. An attacker can exploit this by creating a branch name with a malicious payload and opening a pull request, potentially leading to reverse shell access or theft of sensitive tokens and keys.

A vulnerability in significant-gravitas/autogpt version 0.5.1 allows an attacker to bypass the shell commands denylist settings. The issue arises when the denylist is configured to block specific commands, such as 'whoami' and '/bin/whoami'. An attacker can circumvent this restriction by executing commands with a modified path, such as '/bin/./whoami', which is not recognized by the denylist.

AutoGPT, a component of significant-gravitas/autogpt, is vulnerable to an improper neutralization of special elements used in an OS command ('OS Command Injection') due to a flaw in its shell command validation function. Specifically, the vulnerability exists in versions v0.5.0 up to but not including 5.1.0. The issue arises from the application's method of validating shell commands against an allowlist or denylist, where it only checks the first word of the command. This allows an attacker to bypass the intended restrictions by crafting commands that are executed despite not being on the allowlist or by including malicious commands not present in the denylist. Successful exploitation of this vulnerability could allow an attacker to execute arbitrary shell commands.

An OS command injection vulnerability exists in the MacOS Text-To-Speech class MacOSTTS of the significant-gravitas/autogpt project, affecting versions up to v0.5.0. The vulnerability arises from the improper neutralization of special elements used in an OS command within the `_speech` method of the MacOSTTS class. Specifically, the use of `os.system` to execute the `say` command with user-supplied text allows for arbitrary code execution if an attacker can inject shell commands. This issue is triggered when the AutoGPT instance is run with the `--speak` option enabled and configured with `TEXT_TO_SPEECH_PROVIDER=macos`, reflecting back a shell injection snippet. The impact of this vulnerability is the potential execution of arbitrary code on the instance running AutoGPT. The issue was addressed in version 5.1.0.

A Cross-Site Request Forgery (CSRF) vulnerability in significant-gravitas/autogpt version v0.5.0 allows attackers to execute arbitrary commands on the AutoGPT server. The vulnerability stems from the lack of protections on the API endpoint receiving instructions, enabling an attacker to direct a user running AutoGPT in their local network to a malicious website. This site can then send crafted requests to the AutoGPT server, leading to command execution. The issue is exacerbated by CORS being enabled for arbitrary origins by default, allowing the attacker to read the response of all cross-site queries. This vulnerability was addressed in version 5.1.