Do NOT open a public issue for security vulnerabilities.

Instead, please report security issues via:

1. GitHub Security Advisories: Report a vulnerability
2. Email: Create a private advisory on GitHub (preferred)

• Description of the vulnerability
• Steps to reproduce
• Impact assessment
• Suggested fix (if any)

• Acknowledgment: Within 48 hours
• Assessment: Within 1 week
• Fix: Depends on severity (critical: ASAP, high: 1 week, medium: 2 weeks)

• Each terminal session runs in its own process (forked via PTY)
• Sessions are isolated from each other at the OS level
• The app runs as a regular Android app (no root required)

• Localhost-only binding (127.0.0.1)
• SecurityPolicy blocks dangerous commands (rm -rf /, dd, etc.)
• ApprovalManager gates tool execution
• Disabled by default — user must opt-in via Settings

• Models downloaded from HuggingFace only
• Inference runs entirely on-device (no data leaves the phone)
• LLM is optional — disabled by default
• No telemetry, no analytics, no cloud calls

• Session metadata: app-private JSON (not accessible to other apps)
• Block store: app-private SQLite with WAL mode
• Preferences: app-private SharedPreferences
• Bootstrap: extracted to app-private directory
• No data is sent to any server

• No telemetry or analytics
• No crash reporting to external services
• No network calls except user-initiated (apt, SSH, model download)
• No access to contacts, camera, microphone, or location
• No ads, no tracking