Agentic Tend develops a general agentic context model without encoding intelligence that general models can learn.

Tend means caring for a growing system: preserve non-inferable objectives and preferences, add structure when evidence calls for it, and let projects retain their own shape.

Start from the request, current state, and observed evidence; derive the traits needed to select capabilities, then revise that selection as new evidence appears. A task may compose several capabilities at once; persistent context is added only when future readers or runtimes cannot reliably reconstruct what matters.

• Agentic tooling defines the motivation, task loop, capability composition, and evaluation model.
• Context ownership defines what to persist, where it belongs, and how it is activated or checked.
• Development composition gives concrete examples without prescribing a fixed workflow.
• Skills package reusable capabilities behind task-matching descriptions.
• The organization roadmap tracks evidence and extension work spanning multiple layers.