CVE-2026-58499 in EverOS
Summary
by MITRE • 07/11/2026
EverOS is a memory runtime for agents. Prior to 1.0.1, EverOS is vulnerable to path traversal in the POST /api/v1/memory/add ingestion endpoint because the per-message sender_id field was not validated as a path-safe identifier, unlike app_id and project_id. During user-memory extraction, sender_id is used as owner_id and joined into the filesystem path where the extracted episode is persisted as a Markdown file, so a sender_id containing ../ sequences could direct writes outside the configured memory root and allow an unauthenticated caller to create or overwrite .md files at locations writable by the server process with partially attacker-influenced content. This issue is fixed in version 1.0.1.
If you want to get best quality of vulnerability data, you may have to visit VulDB.
Analysis
by VulDB Data Team • 07/11/2026
EverOS represents a sophisticated memory runtime environment designed for agent-based systems, where the platform manages user memory fragments through a structured ingestion and persistence mechanism. The vulnerability resides within the POST /api/v1/memory/add endpoint which serves as the primary interface for adding memory content to the system. This endpoint processes incoming messages containing various identifiers including sender_id, app_id, and project_id fields that are essential for organizing and storing memory fragments within the filesystem.
The core technical flaw manifests in the inadequate validation of the sender_id field, which creates a critical path traversal vulnerability. While app_id and project_id fields undergo proper sanitization to ensure they contain only path-safe characters, the sender_id field remains unchecked and unvalidated against directory traversal sequences. This inconsistent validation approach allows attackers to inject malicious path components such as ../ into the sender_id parameter. The vulnerability becomes particularly dangerous during the memory extraction process where the attacker-influenced sender_id value is directly used as the owner_id identifier and subsequently incorporated into filesystem paths for Markdown file creation.
During the operational execution of this vulnerability, an unauthenticated attacker can exploit the flaw by crafting a malicious request with a sender_id containing directory traversal sequences. When the system processes this request, it constructs filesystem paths using the unchecked sender_id value, effectively allowing the attacker to write or overwrite .md files outside the intended memory root directory. The impact extends beyond simple file creation as attackers can potentially overwrite existing system files or create malicious content in locations writable by the server process, leading to potential privilege escalation or data corruption scenarios.
The vulnerability directly maps to CWE-22 Path Traversal and aligns with ATT&CK technique T1059 Command and Scripting Interpreter for executing arbitrary file operations. The issue demonstrates poor input validation practices where different fields within the same API endpoint receive inconsistent security treatments, creating an exploitable gap in the system's defenses. The fix implemented in version 1.0.1 addresses this by introducing proper validation of the sender_id field to ensure it contains only safe path characters and rejects any sequences that could lead to directory traversal operations, thereby maintaining consistency with the validation applied to app_id and project_id fields.
This vulnerability highlights the importance of consistent security controls across all input parameters within API endpoints and demonstrates how seemingly minor oversights in field validation can lead to significant operational risks. The remediation approach emphasizes defensive programming principles where all user-controllable inputs undergo identical validation treatments, preventing attackers from leveraging inconsistent security measures to gain unauthorized access to system resources or manipulate stored memory fragments.