[Roadmap] Yana Autonomous Runtime — current execution plan

[yanacuti1121]

Goal

Move Yana from policy-aware automation to a production closed loop where routine L0-L3 work advances automatically and only sovereign L4 decisions require the human owner.

Current foundation

• PR feat(os): add native system health monitoring #165 — native cross-platform system-health monitoring.
• PR feat(os): add autonomy ladder and durable action queue #166 — L0-L4 autonomy policy and durable action-intent queue.

Execution order

• [Yana OS] Build the typed allowlisted autonomy executor #168 — typed allowlisted autonomy executor.
• [Yana OS] Add automatic scheduler and cross-platform process supervisor #169 — automatic scheduler and cross-platform process supervisor.
• [Yana OS] Close the verification, rollback, and audit loop #170 — verification, rollback, and audit loop.
• [Validation] Prove autonomy and monitor behavior on macOS, Linux, and Windows #172 — macOS/Linux/Windows end-to-end validation.
• [UX] Surface system health and autonomy state across Yana interfaces #171 — unified terminal/web/desktop health and autonomy status.

Sovereign boundary

Yana may automatically observe, diagnose, apply verified reversible fixes, create bounded development artifacts, and prepare draft review work. It must stop for explicit human approval before protected-branch merge, release, production deployment, secret rotation, destructive persistent-data changes, or security-policy changes.

Completion criteria

• Automatic actions are selected only through typed capability policy.
• Crashes cannot duplicate or silently lose queued work.
• Every mutation has verification evidence and an applicable rollback path.
• Cross-platform claims are backed by real CI/runtime evidence.
• Terminal and desktop surfaces report truthful health and queue state.
• No path bypasses yana-rt, guardrails, audit evidence, or sovereign gates.

Attribution

Roadmap owner: Vũ Văn Tâm. Initial issue structure: OpenAI Codex.