The knowledge unit that survives agent decomposition.
Six mandatory faces. Quality-gated at 42/60. Agent-parseable from birth.
Every cubelet has six mandatory faces. Every face scores minimum 6/10. Aggregate hits 42/60. No face compensates for another.
Business stakes — why this knowledge matters in context.
Definition + boundaries — precise scope of the concept.
Mechanism — how it works under the surface.
Context + limits — when it applies and when it doesn't.
Triggers + workflow — the conditions that activate it.
Step-by-step framework + worked example.
Quality gate
No blank faces allowed. Every face scores minimum 6/10. The aggregate must hit 42/60. No face compensates for another — a perfect APPLY score cannot mask a missing WHY.
Signal to ship in 5 days. Five specialized agents, each with a defined role.
Gathers source material and domain context.
Drafts all six faces from research output.
Refines clarity, accuracy, and completeness.
Scores every face. Enforces 42/60 quality gate.
Dual-serializes and publishes to substrate.
Human-readable. Agent-parseable.
Every cubelet is born in two forms. Humans read structured content. AI agents consume JSON with typed fields. Same knowledge, two interfaces, no translation loss.
Browse cubelets →