About

Under stated assumptions there is a structure-preserving interpretation of the four pillars of agent externalization, memory, skills, protocols and harness, in the ArchAgents Architecture triple (G, Know, Phi).

Three production systems realise it on identifiable fragments: ContextFS for memory, AgentHero for skills and for the harness control plane, and CatDB for protocols. Each part of the series fixes one pillar against one system at a pinned commit and states which conditions hold, which hold only on a restricted fragment, and which fail.

The series states where the interpretation stops working as plainly as where it works, with a witness in the source for every clause that fails.

The four pillars

Pillar	Categorical role	System	Part
Memory	Lineage coalgebra over the wiring G	ContextFS	Part I
Skills	Operations of a coloured operad on G	AgentHero	Part II
Protocols	Typed wiring G itself, with refusal-complete ports	CatDB	Part III
Harness	The triple (G, Know, Phi) and its monoidal functors	AgentHero control plane	Part IV

Framing papers

The series tests two prior arguments against production systems rather than restating them: Banu’s Architecture triple (arXiv:2605.12239) and Cao’s account of agentic restructuring (arXiv:2606.05608). Source PDFs of both are linked to arXiv, not redistributed here.

Author

Matthew Long, The YonedaAI Collaboration, YonedaAI Research Collective, Chicago, IL. matthew@yonedaai.com · yonedaai.com