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.

• Start with the Agentic Tend guide to see how Human authority, LLM reasoning, and external evidence form the interaction loop.
• The organization documentation map routes questions about behavior admission, capability dispatch, context persistence, human review, evidence separation, and context evaluation to their canonical owners. Capability-composition examples remain a derived view.
• Skills package reusable capabilities behind task-matching descriptions.