Skip to content

Security: suterkl74014/LinuxTerminalAgent

Security

SECURITY.md

Security policy and operational warning

Privilege model

Linux Terminal Agent is intentionally powerful. The WSL-side shell runs as root, Windows drives mounted under /mnt are real host filesystems, and optional desktop/browser helpers can manipulate the Windows session. Treat agent instructions and third-party content as potentially unsafe input.

Recommended operating practices

  • Keep snapshots or normal backups before allowing broad filesystem changes.
  • Use a dedicated working folder and inspect destructive commands.
  • Disable Windows desktop control when it is not required.
  • Do not commit .env, runtime secret files, .lta-keys, legacy .agentsmith-keys, logs, or provider credentials.
  • Use provider-scoped keys with the least privilege the provider supports.
  • Keep the localhost control token private; do not expose local helper ports through port-forwarding or firewall exceptions.

Secret handling

The application stores provider keys locally and redacts them from ordinary settings/status output. API-key backup files are encrypted, but the backup file and its password should be stored separately. ChatGPT/Codex account sessions are not included in API-key backups.

Reporting a vulnerability

If this project is published on GitHub, configure a private security-reporting channel for the repository and place its address/instructions here. Do not post live API keys, tokens, private logs, or exploit details in public issues.

There aren't any published security advisories