CVE-2026-7384
Description
A vulnerability was detected in ezequiroga mcp-bases 357ca19c7a49a9b9cb2ef639b366f03aba8bea39/c630b8ab0f970614d42da8e566e9c0d15a16414c. This impacts the function search_papers of the file research_server.py. Performing a manipulation of the argument topic results in path traversal. Remote exploitation of the attack is possible. The exploit is now public and may be used. This product follows a rolling release approach for continuous delivery, so version details for affected or updated releases are not provided. 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.
Path traversal in ezequiroga mcp-bases research_server.py allows remote attackers to write JSON files outside the intended papers/ directory via the topic parameter.
Vulnerability
Overview
The search_papers function in research_server.py of the ezequiroga mcp-bases project contains a path traversal vulnerability (CWE-22). The function accepts a user-controlled topic argument and constructs a filesystem path by joining the papers/ directory with the lowercased, space-replaced topic string. No canonicalization or boundary check is applied, so traversal sequences like ../ are preserved. This allows an attacker to escape the intended storage directory [1].
Exploitation
An attacker can send a crafted topic parameter containing ../ sequences to the MCP server. The server will create the corresponding directory (e.g., papers/../escape-test) and later write a papers_info.json file into that escaped location. The resource handler papers://{topic} uses the same unsafe path construction, enabling reading of files outside the intended scope. The exploit does not require authentication, as the MCP server is typically exposed to clients [1]. The exploit code has been made public, increasing the immediate risk [1].
Impact
Successful exploitation allows an attacker to create arbitrary directories and write JSON files anywhere the server process has write permissions. This could lead to overwriting configuration files, planting malicious data, or other unauthorized file operations. The project's rolling release model means no specific version numbers are provided, but the vulnerability exists in the scanned revision and likely in all revisions with the same path construction [1][2].
Mitigation
As of the publication date, the vendor has not responded to the issue report and no patch is available [1]. Users should monitor the project repository for updates. As a workaround, restrict network access to the MCP server to trusted clients only, or implement input validation to reject path traversal sequences in the topic parameter [1][2].
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.
Affected products
1- Range: at commit 357ca19c7a49a9b9cb2ef639b366f03aba8bea39 (or c630b8ab0f970614d42da8e566e9c0d15a16414c)
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.