CVE-2026-42075
Published: 04 May 2026
Description
Evolver is a GEP-powered self-evolving engine for AI agents. Prior to version 1.69.3, a path traversal vulnerability in the skill download (fetch) command allows attackers to write files to arbitrary locations on the filesystem. The --out= flag accepts user-provided paths…
more
without validation, enabling directory traversal attacks that can overwrite critical system files or create files in sensitive location. This issue has been patched in version 1.69.3.
Mitigating Controls (NIST 800-53 r5)AI
Directly mandates validation of user-provided paths in the --out= flag to block directory traversal sequences like ../.
Requires identification, reporting, and patching of the specific path traversal flaw fixed in Evolver version 1.69.3.
Enforces least privilege to restrict low-privileged (PR:L) attackers from writing to or overwriting critical system files outside intended directories.
Security SummaryAI
CVE-2026-42075 is a path traversal vulnerability (CWE-22) affecting Evolver, a GEP-powered self-evolving engine for AI agents, in versions prior to 1.69.3. The issue resides in the skill download (fetch) command, where the --out= flag accepts user-provided paths without validation. This allows attackers to specify paths that traverse directories, enabling writes to arbitrary locations on the filesystem, such as overwriting critical system files or creating files in sensitive directories. The vulnerability carries a CVSS v3.1 base score of 8.1 (AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:H).
Attackers with low privileges (PR:L) can exploit this over the network (AV:N) with low complexity and no user interaction required. By providing a malicious path to the --out= flag during execution of the fetch command, such as one using directory traversal sequences (e.g., ../), they can achieve arbitrary file writes. This results in high integrity (I:H) and availability (A:H) impacts, potentially disrupting system operations or escalating control through file manipulation, though no confidentiality impact is present.
The vulnerability has been patched in Evolver version 1.69.3, as detailed in the project's GitHub release notes and security advisories (GHSA-r466-rxw4-3j9j). Security practitioners should upgrade to 1.69.3 or later to mitigate the issue, and review usage of the fetch command's --out= flag in automated or user-controlled environments.
Evolver's role as a self-evolving engine for AI agents introduces relevance to AI/ML deployments, where untrusted skill downloads could amplify risks in agentic workflows. No public evidence of real-world exploitation is available as of the CVE publication on 2026-05-04.
Details
- CWE(s)
AI Security AnalysisAI
- AI Category
- Other AI Platforms
- Risk Domain
- N/A
- OWASP Top 10 for LLMs 2025
- None mapped
- MITRE ATLAS Techniques
- None mapped
- Classification Reason
- Matched keywords: ai
MITRE ATT&CK Enterprise TechniquesAI
Why these techniques?
Path traversal in the skill fetch --out= flag provides an arbitrary file write primitive on the target filesystem. This directly enables placement of attacker-controlled files for tool ingress (T1105), writing cron/systemd/scheduled-task definitions (T1053), creation or modification of persistent system processes/services (T1543), and overwriting security-tool binaries or configuration files (T1562).