CVE-2026-75130 describes a prompt injection vulnerability in Uptash's Context7 product through version 2.1.2. The vulnerability arises from improper neutralization of input used for AI prompt generation in the Custom AI Instructions feature served by the MCP server. Attackers can inject unsanitized content that causes connected AI coding agents to execute malicious instructions. This can result in exfiltration of credentials from environment files to attacker-controlled services and destructive file deletions on the victim's machine during normal agent operations such as library documentation requests.

Successful exploitation allows attackers to execute arbitrary instructions within AI coding agents, leading to credential theft and destructive file operations on the victim's system. This compromises confidentiality and integrity of the affected environment.

No patch or official fix information is provided in the available data. Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until a fix is available, restrict access to the Custom AI Instructions feature and monitor for suspicious activity related to AI agent operations.