Skip to content

Security: Q-YZX0/Q-Agent

Security

SECURITY.md

Security Policy

Supported versions

This is an active project. Security fixes are applied to main and backported as needed.

Reporting a vulnerability

  • Please open a private report via email or a confidential issue (do not post details publicly).
  • Include: steps to reproduce, impact, affected area (core or module), and suggested remediation if any.

Threat model

  • Local assistant with system control (Windows-first). Primary risks:
    • Privilege misuse: filesystem/process/network/UI automation
    • Supply chain: third-party modules
    • Data leakage: logs, memory, secrets

Protections in core

  • Capabilities and permissions per module: filesystem.read|write, process.spawn, network.request, ui.automation
  • Permission prompts with time-bound elevation; whitelist in core/config/qai_config.json
  • Security policy: security.allow/deny per command, path_limits to restrict filesystem operations
  • Audit trail: command executions stored in memory; rotating logs under logs/
  • Timeouts/retries and isolation of dynamic commands execution via thread pool
  • Health checks and safe degradation to text mode if voice deps are missing

Modules and marketplace

  • Manifest v1 (id, version, permissions, commands, optional ui.json)
  • Dynamic loader: loads only declared commands/handlers, records manifest metadata
  • Recommended: run third-party modules with restricted permissions and review code before enabling

Secrets and logs

  • Store secrets in environment variables (.env ignored by VCS)
  • Do not include tokens in code or UI schemas
  • Logs redact sensitive fields where possible; rotate regularly

UI automation

  • ui.automation is sensitive: requires explicit opt-in (whitelist) and is time-bound when prompted
  • Prefer path-limited operations and confirmation for high-impact actions

Responsible disclosure

We appreciate responsible disclosure. You will receive acknowledgment in the changelog after the fix is released.

There aren’t any published security advisories