LEGAL-03 // SECURITY_POLICY

Security Policy

Responsible disclosure, threat model, and the five mitigations built into every release.

THREAT MODEL // FIVE MITIGATIONS

Five threats. Five mitigations.

[X]LLM hallucination → rm -rf /

[✓]LLM output is NEVER executed on the host directly — always via Docker sandbox

[X]Daemon privilege escalation

[✓]Daemon runs as yantra_daemon (UID 999) with strict systemd sandboxing

[X]Container escape to host filesystem

[✓]Docker has NO --network and NO host mount — stdout only

[X]Unrestricted SSH command execution

[✓]SSH key allows ONLY whitelisted commands (systemctl, pacman, fstrim…)

[X]Secrets exfiltration via prompt injection

[✓]Secrets in /etc/yantra/secrets.env (chmod 0400) — never in model context

RESPONSIBLE DISCLOSURE

Found a vulnerability?

Please do not open a public GitHub issue for security vulnerabilities. Email us at security@euryaleferox.com with:

01.A description of the vulnerability and affected component

02.Steps to reproduce, including environment details

03.Your assessment of severity and potential impact

04.Any proof-of-concept code (keep it responsible)

We acknowledge reports within 48 hours and aim to patch critical issues within 14 days. We do not currently offer a bug bounty program, but we will credit you in the Changelog.

SCOPE // BOUNDARY

What's in scope.

In scope

›Kriya Loop daemon (core/daemon.py)
›Docker sandbox executor (core/sandbox.py)
›TUI shell IPC (core/tui_shell.py)
›OTA manager (core/ota_manager.py)
›yantraos.com web interface
›Skill Store API endpoints

Out of scope

›Third-party dependencies (Ollama, Pinecone, Supabase)
›The underlying Arch Linux packages
›User-published Skills
›Network infrastructure not operated by us