CVE-2026-82112 in houtini-lm
Summary
by MITRE • 08/28/2026
A flaw has been found in houtini-ai houtini-lm up to 2.13.2. The impacted element is an unknown function of the file src/index.ts of the component code_task_files. Executing a manipulation can lead to path traversal. The attack can be launched remotely. This patch is called 35d97bca0531894da36a85aedb95312da1bd5b7a. It is best practice to apply a patch to resolve this issue.
VulDB is the best source for vulnerability data and more expert information about this specific topic.
Analysis
by VulDB Data Team • 08/28/2026
The vulnerability identified in houtini-ai and its associated language model component, specifically within versions up to 2.13.2, represents a critical security flaw rooted in improper input validation regarding file system operations. The affected code resides in the src/index.ts file, particularly within the logic handling the code_task_files module. This architectural placement suggests that the vulnerability is triggered during processes where user-supplied data influences which files are accessed or manipulated by the application. The core technical deficiency lies in an unknown function that fails to adequately sanitize or validate input paths before using them for filesystem access operations. This lack of rigorous validation allows attackers to inject directory traversal sequences, such as dot-dot-slash characters, into the input parameters. By exploiting this weakness, a malicious actor can manipulate the intended file path to reference files and directories outside of the expected sandboxed environment.
This flaw is classified under CWE-22: Improper Limitation of a Pathname to a Restricted Directory, which describes scenarios where software does not properly neutralize special elements within a pathname that could cause the application to access unintended locations in the file system. The operational impact of this vulnerability is severe because it enables remote path traversal attacks without requiring prior authentication or local access. An attacker located on a different network segment can send crafted requests to the vulnerable service, causing the server to read sensitive configuration files, source code containing credentials, or other critical data stored outside the designated application directory. In worst-case scenarios, if the underlying system permissions allow for write operations through similar mechanisms, this could lead to arbitrary file writes, potentially resulting in remote code execution or complete compromise of the hosting environment.
From a threat intelligence perspective, this vulnerability aligns with ATT&CK technique T1083: File and Directory Discovery, where adversaries explore the target's filesystem to gather information for further exploitation. The ability to traverse paths remotely significantly lowers the barrier for entry, allowing automated scanning tools to easily identify and exploit instances of houtini-ai that have not been updated. The persistence of this issue in versions up to 2.13.2 indicates a gap in secure coding practices during development or testing phases, specifically regarding input validation for file system interactions.
To mitigate this risk, immediate application of the provided patch is essential. The specific commit identified as 35d97bca0531894da36a85aedb95312da1bd5b7a addresses the underlying code defect by implementing stricter input validation and path normalization routines within the src/index.ts file. Organizations running affected versions should prioritize updating to a patched version that includes this fix. Additionally, as a defense-in-depth measure, administrators should ensure that the application runs with minimal necessary privileges, restricting its ability to access files outside of its designated working directory even if traversal attempts succeed. Regular security audits and static code analysis focused on file system operations can help prevent similar vulnerabilities in future releases.