Kernel.
Capabilities.
Cells.
KCC exists to structurally compound intelligence without collapsing governance — an AI-native engineering operating model that turns agentic work into contracts, traces, reusable capabilities, human gates, and verifiable delivery.
What KCC answers
The model answers the practical questions teams keep running into with AI coding assistants — context engineering, subagents and skills, supervision level, sandboxing, security, cost, maintainability, verification, and feedback loops.
"What is the state of play for AI coding assistants — trust, supervision, and practice?"
"Can an LLM interview the human to surface hidden requirements before answering?"
"What makes an agent dangerous — private data, untrusted content, and external action?"
Layered entry points
Browse the whole model
Introduction
Core Thesis
Foundations
Agent Contract
Cognitive Layer
Inspection
Lifecycle
Special Concerns
Reference Examples
Boundaries
Run the doctrine, don't just read it.
The KCC agentic framework packages this operating model as harness-neutral agents and slash-command skills — onboard, interrogate, spec, budget, plan, implement, verify — with the human gate wherever confidence or budget says stop.