CVE-2026-66065: Ouroboros arbitrary command execution through untrusted project .env keys

Summary

Versions of Ouroboros, a local-first runtime for AI coding agents, prior to 0.42.1 ship an incomplete denylist of untrusted .env keys. A malicious cloned repository can include a .env file that is auto-loaded at import with no review step, letting an attacker reach arbitrary command execution. The earlier CVE-2026-47211 fix added _UNTRUSTED_ENV_DENYLIST for project-directory .env files but missed several keys. Other variables let attackers redirect the backend config-home and MCP/plugin roots past the approval gate, re-enable blocked local transports, replace sub-agent prompts, and lower tool approval classes.

Mitigation

Fixed in 0.42.1.

Topics

• AI agents

Related items

• MediumCVE-2026-121936: easyVoice server-side request forgery through openaiBaseUrl in /api/v1/ttsSimilar attack · NVD/CVE Database
• MediumCVE-2026-118628: MLflow get-artifacts server-side request forgery information disclosureSimilar attack · NVD/CVE Database
• HighCVE-2026-118625: MLflow Tracking Server privilege escalation via get_endpointsSimilar attack · NVD/CVE Database
• InfoWhen trust becomes the payload: Inside the threats, etc quarterly reportSimilar attack · Elastic Security Labs
• MediumCritical Langflow flaws enable authentication bypass and RCESimilar attack · CSO Online