PraisonAI: Critical RCE via `type: job` workflow YAML
PraisonAI is a multi-agent teams system. In versions below 4.5.139 of PraisonAI and 1.5.140 of praisonaiagents, the workflow engine is vulnerable to arbitrary command and code execution through untrusted YAML files. When praisonai workflow run <file.yaml> loads a YAML file with type: job, the JobWorkflowExecutor in job_workflow.py processes steps that support run: (shell commands via subprocess.run()), script: (inline Python via exec()), and python: (arbitrary Python script execution)—all without any validation, sandboxing, or user confirmation. The affected code paths include action_run() in workflow.py and _exec_shell(), _exec_inline_python
PraisonAI is a multi-agent teams system. Versions 4.5.138 and below are vulnerable to arbitrary code execution through automatic, unsanitized import of a tools.py file from the current working directory. Components including call.py (import_tools_from_file()), tool_resolver.py (_load_local_tools()), and CLI tool-loading paths blindly import ./tools.py at startup without any validation, sandboxing, or user confirmation. An attacker who can place a malicious tools.py in the directory where PraisonAI is launched (such as through a shared project, cloned repository, or writable workspace) achieves immediate arbitrary Python code execution in the
PraisonAI: SQLiteConversationStore didn't validate table_prefix when constructing SQL queries
PraisonAI is a multi-agent teams system. Prior to 4.5.133, there is an SQL identifier injection vulnerability in SQLiteConversationStore where the table_prefix configuration value is directly concatenated into SQL queries via f-strings without any validation or sanitization. Since SQL identifiers cannot be safely parameterized, an attacker who controls the table_prefix value (e.g., through from_yaml or from_dict configuration input) can inject arbitrary SQL fragments that alter query structure. This enables unauthorized data access, such as reading internal SQLite tables like sqlite_master, and manipulation of query results through techniques
Средний · CVSS 4.3EPSS 0,29%Нет опубликованного исправления
Open WebUI has Blind Server Side Request Forgery in its Image Edit Functionality
Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Versions 0.7.2 and below contain a Blind Server Side Request Forgery in the functionality that allows editing an image via a prompt. The affected function performs a GET request to a user-provided URL with no restriction on the domain, allowing the local address space to be accessed. Since the SSRF is blind (the response cannot be read), the primary impact is port scanning of the local network, as whether a port is open can be determined based on whether the GET request succeeds or fails. These response differentials can be automated to iterate
Incomplete mitigation of CVE-2026-4519, %action expansion for command injection to webbrowser.open()
Mitgation of CVE-2026-4519 was incomplete. If the URL contained "%action" the mitigation could be bypassed for certain browser types the "webbrowser.open()" API could have commands injected into the underlying shell. See CVE-2026-4519 for details.
Use-after-free in lzma.LZMADecompressor, bz2.BZ2Decompressor, and gzip.GzipFile after re-use under memory pressure
Use-after-free (UAF) was possible in the `lzma.LZMADecompressor`, `bz2.BZ2Decompressor`, and `gzip.GzipFile` when a memory allocation fails with a `MemoryError` and the decompression instance is re-used. This scenario can be triggered if the process is under memory pressure. The fix cleans up the dangling pointer in this specific error condition.
The vulnerability is only present if the program re-uses decompressor instances across multiple decompression calls even after a `MemoryError` is raised during decompression. Using the helper functions to one-shot decompress data such as `lzma.decompress()`, `bz2.decompress()`, `gzip.decompress()`,
A vulnerability in the `TFSMLayer` class of the `keras` package, version 3.13.0, allows attacker-controlled TensorFlow SavedModels to be loaded during deserialization of `.keras` models, even when `safe_mode=True`. This bypasses the security guarantees of `safe_mode` and enables arbitrary attacker-controlled code execution during model inference under the victim's privileges. The issue arises due to the unconditional loading of external SavedModels, serialization of attacker-controlled file paths, and the lack of validation in the `from_config()` method.
Apache Airflow: Unsafe Deserialization via Legacy Serialization Keys (__type/__var) Bypass in XCom API
Dag Authors, who normally should not be able to execute code in the webserver context could craft XCom payload causing the webserver to execute arbitrary code. Since Dag Authors are already highly trusted, severity of this issue is Low.
Users are recommended to upgrade to Apache Airflow 3.2.0, which resolves this issue.
Apache Airflow: Secrets from Airflow config file logged in plain text in DAG run logs UI
Before Airflow 3.2.0, it was unclear that secure Airflow deployments require the Deployment Manager to take appropriate actions and pay attention to security details and security model of Airflow. Some assumptions the Deployment Manager could make were not clear or explicit enough, even though Airflow's intentions and security model of Airflow did not suggest different assumptions. The overall security model [1], workload isolation [2], and JWT authentication details [3] are now described in more detail. Users concerned with role isolation and following the Airflow security model of Airflow are advised to upgrade to Airflow 3.2, where several
Remote Code Execution in Google Agent Development Kit (ADK)
A Code Injection and Missing Authentication vulnerability in Google Agent Development Kit (ADK) versions 1.7.0 (and 2.0.0a1) through 1.28.1 (and 2.0.0a2) on Python (OSS), Cloud Run, and GKE allows an unauthenticated remote attacker to execute arbitrary code on the server hosting the ADK instance.
This vulnerability was patched in versions 1.28.1 and 2.0.0a2.
Customers need to redeploy the upgraded ADK to their production environments. In addition, if they are running ADK Web locally, they also need to upgrade their local instance.
A security flaw has been discovered in FoundationAgents MetaGPT up to 0.8.1. This impacts the function decode_image of the file metagpt/utils/common.py. The manipulation of the argument img_url_or_b64 results in server-side request forgery. It is possible to launch the attack remotely. The exploit has been released to the public and may be used for attacks. The project was informed of the problem early through an issue report but has not responded yet.
A Cross-site Scripting (XSS) vulnerability was identified in the `from_dict` method of the `AppLollmsMessage` class in parisneo/lollms prior to version 2.2.0. The vulnerability arises from the lack of sanitization or HTML encoding of the `content` field when deserializing user-provided data. This allows an attacker to inject malicious HTML or JavaScript payloads, which can be executed in the context of another user's browser. Exploitation of this vulnerability can lead to account takeover, session hijacking, or wormable attacks.
A vulnerability was identified in FoundationAgents MetaGPT up to 0.8.1. This affects the function generate_thoughts of the file metagpt/strategy/tot.py of the component Tree-of-Thought Solver. The manipulation leads to code injection. It is possible to initiate the attack remotely. The exploit is publicly available and might be used. The project was informed of the problem early through an issue report but has not responded yet.
Средний · CVSS 5.3EPSS 0,22%Нет опубликованного исправления
FoundationAgents MetaGPT Mineflayer HTTP API index.js evaluateCode cross-site request forgery
A vulnerability was determined in FoundationAgents MetaGPT up to 0.8.1. The impacted element is the function evaluateCode of the file metagpt/environment/minecraft/mineflayer/index.js of the component Mineflayer HTTP API. Executing a manipulation can lead to cross-site request forgery. The attack may be performed from remote. The exploit has been publicly disclosed and may be utilized. The project was informed of the problem early through an issue report but has not responded yet.
aws-mcp-server AWS CLI Command Injection Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of aws-mcp-server. Authentication is not required to exploit this vulnerability.
The specific flaw exists within the handling of the allowed commands list. The issue results from the lack of proper validation of a user-supplied string before using it to execute a system call. An attacker can leverage this vulnerability to execute code in the context of the MCP server. Was ZDI-CAN-27969.
rembg server is vulnerable to Server-Side Request Forgery (SSRF) and a weak default CORS configuration
# GitHub Security Lab (GHSL) Vulnerability Report, rembg: `GHSL-2024-161`, `GHSL-2024-162`
The [GitHub Security Lab](https://securitylab.github.com) team has identified potential security vulnerabilities in [rembg](https://github.com/danielgatis/rembg).
We are committed to working with you to help resolve these issues. In this report you will find everything you need to effectively coordinate a resolution of these issues with the GHSL team.
If at any point you have concerns or questions about this process, please do not hesitate to reach out to us at `securitylab@github.com` (please include `GHSL-2024-161` or `GHSL-2024-162` as a reference
GeoNode versions 4.4.5 and 5.0.2 (and prior within their respective releases) contain a server-side request forgery vulnerability in the service registration endpoint that allows authenticated attackers to trigger outbound network requests to arbitrary URLs by submitting a crafted service URL during form validation. Attackers can probe internal network targets including loopback addresses, RFC1918 private IP ranges, link-local addresses, and cloud metadata services by exploiting insufficient URL validation in the WMS service handler without private IP filtering or allowlist enforcement.
GeoNode versions 4.0 before 4.4.5 and 5.0 before 5.0.2 contain a server-side request forgery vulnerability that allows authenticated users with document upload permissions to trigger arbitrary outbound HTTP requests by providing a malicious URL via the doc_url parameter during document upload. Attackers can supply URLs pointing to internal network targets, loopback addresses, RFC1918 addresses, or cloud metadata services to cause the server to make requests to internal resources without SSRF mitigations such as private IP filtering or redirect validation.
xrootd has path traversal in directory listing that allows access to the parent directory via trailing ".." pattern
## Summary
A path traversal vulnerability in XRootD allows users to escape the exported directory scope and enumerate the contents of the parent directory by appending `/..` (specifically without trailing slash) to an exported path in `xrdfs ls` or `HTTP PROPFIND` requests.
This bypass ignores the `all.export` restriction.
## Affected component
`src/XrdXrootd/XrdXrootdXeq.cc`, and more precisely the functions `rpCheck()` and `Squash()` used in `do_Dirlist()` ([link](https://github.com/xrootd/xrootd/blob/19aa6dee76906fb4d56ded55e49bbe4171ade915/src/XrdXrootd/XrdXrootdXeq.cc#L696)), as they do not check if the path ends with ".." (without t
ajenti.plugin.core defines all necessary core elements to allow Ajenti to run properly. Prior to 0.112, if the 2FA was activated, it was possible during a short moment after the authentication of an user to bypass its authentication. This vulnerability is fixed in 0.112.