Autonomous Self-Management Cycles

Knows What’s Wrong Before You Do · Shipped · workstation, home

5-phase background loops (Sense → Deliberate → Propose → Act → Reflect) driving autonomous issue resolution.

A somatic block is the wrapper that drives a single self-management cycle through five phases. It references — but does not absorb — the existing Finding, Proposal, ApprovalRequest, Action, and Reflection models by ID, composing them into a complete autonomous resolution arc.

The SomaticLifecycle orchestrator drives each block through its phases: SENSORY gathers scanner data, DELIBERATION evaluates severity and blast radius, PROPOSAL generates a config change through the ProposalGenerator, ACTION executes through the approval engine (if approved), and REFLECTION logs the outcome and adjusts future detection thresholds.

Limits and invariants

Every action phase routes through the approval engine — somatic blocks cannot bypass human-in-the-loop authorization for state-mutating operations.

Where this lives

• halbert_core/halbert_core/somatic/lifecycle.py
• halbert_core/halbert_core/somatic/block.py
• halbert_core/halbert_core/somatic/store.py