CVE-2026-7235
Description
A security vulnerability has been detected in ErlichLiu claude-agent-sdk-master up to b185aa7ff0d864581257008077b4010fca1747bf. Affected by this vulnerability is an unknown functionality of the file app/api/agent-output/route.ts. The manipulation of the argument outputFile leads to path traversal. The attack may be initiated remotely. The exploit has been disclosed publicly and may be used. This product uses a rolling release model to deliver continuous updates. As a result, specific version information for affected or updated releases is not available. The project was informed of the problem early through an issue report but has not responded yet.
AI Insight
LLM-synthesized narrative grounded in this CVE's description and references.
A path traversal vulnerability in ErlichLiu claude-agent-sdk-master allows remote attackers to read arbitrary local files via the outputFile argument in /api/agent-output.
A path traversal vulnerability has been discovered in the ErlichLiu claude-agent-sdk-master project, specifically in the 04-agent-teams component within the file app/api/agent-output/route.ts [1]. The issue is classified as CWE-22 (Improper Limitation of a Pathname to a Restricted Directory) [2]. The vulnerability stems from the endpoint accepting a user-supplied outputFile value from the request body and passing it directly to fs.readFile after path normalization, without verifying that the resolved path stays within a trusted agent output directory or application workspace [2].
An attacker with network access to the exposed Next.js API endpoint can exploit this by manipulating the outputFile argument to traverse directories and read arbitrary files readable by the server process [2]. The attack can be initiated remotely without authentication beyond network access to the endpoint [1]. The vulnerability has been publicly disclosed along with details of the root cause, increasing the risk of active exploitation [2].
Successful exploitation allows an attacker to read sensitive information from the server, which may include configuration files, credentials, source code, or other confidential data stored on the filesystem [2]. This could lead to further compromise of the application or underlying infrastructure if secrets or private keys are exposed. The project uses a rolling release model and specific version information is not available, but the confirmed affected commit is b185aa7ff0d864581257008077b4010fca1747bf [1][2].
As of the publication date, no fix has been released, and the project has not responded to the initial report [1]. Users of affected versions should either apply input validation to restrict the outputFile path to a predefined allowed directory or disable the endpoint until a patch is issued [2]. The vulnerability has been assigned a CVSS v3 base score of 5.3 (Medium) [1].
- GitHub - ErlichLiu/claude-agent-sdk-master: Learn to build real-world applications with Claude Agent SDK. This tutorial takes you from basics to advanced implementations, covering essential integrations (Skills, MCP) and the core principles of human-AI interaction.
- Arbitrary Local File Read Vulnerability in 04-agent-teams of claude-agent-sdk-master
AI Insight generated on May 18, 2026. Synthesized from this CVE's description and the cited reference URLs; citations are validated against the source bundle.
Patches
0No patches discovered yet.
Vulnerability mechanics
AI mechanics synthesis has not run for this CVE yet.
References
4News mentions
0No linked articles in our index yet.