EA Agent Mesh — Operating Model

A multi-cadence agent mesh that runs Enterprise Architecture as a continuously operating practice — five clock speeds (Daily / Weekly / Monthly / Quarterly / Yearly), twelve core artefacts, thirteen agents, one orchestrator.

See /docs/whitepaper for the executive rationale, /docs/deck for the board-level summary, and /docs/adr for the architecture decisions behind this repo's structure.

Repository map

Path	What lives here	Owned by
docs/	Whitepaper, deck, visualization, ADRs	EA practice lead
artefacts/	The 12 EA artefacts as living, versioned data (not slideware)	Per-artefact agent + human approver
mesh/	Orchestrator + 13 agents + policy-as-code + connectors	Engineering
cadence/	Scheduling/trigger config for the 5 loops	Orchestrator owner
stakeholder-views/	Per-role dashboard/digest configs	Platform team
assets/	Logo and favicon files served by the published site	Brand/design owner
(external) Getamazednow AI Design System v1.2/	Brand tokens, templates and logo files. Vendored locally, deliberately not committed — this repo is public and the design system is not published with it. Obtain it from the design-system master repo and place it at the project root before producing artefacts. See Design-System-Localisation.md.	Brand/design owner
Design-System-Localisation.md	Project-specific localisation of the design system (artefact names, version/status, colour remap)	Brand/design owner
infra/	Deployment IaC for the mesh	Platform team
tests/	Unit/integration tests for agents and policy gates	Engineering

The five loops

Daily (Sense) → Weekly (Review) → Monthly (Adjust) → Quarterly (Realign) → Yearly (Reset). Signal flows up, constraints flow down. See the whitepaper §3 for full detail.

Adoption phasing

This repo is intended to be built out bottom-up, matching the whitepaper's adoption roadmap: Phase 1 Daily+Weekly → Phase 2 +Monthly → Phase 3 +Quarterly → Phase 4 +Yearly. Do not scaffold agents ahead of the phase you are actually running.

Repo automation

Governance isn't a document here — it's enforced by two scripts, both driven by the single source of truth config/repo-governance.yml.

Script	When	What it does
scripts/bootstrap-repo.sh	Post-push — once after your first git push, and again whenever config/repo-governance.yml changes	Applies branch protection (required status checks + CODEOWNERS review) to main, creates/updates labels, and checks that every team referenced in CODEOWNERS actually exists — it warns rather than silently creating people-teams
scripts/pre-pr-check.sh	Pre-PR — run locally before opening a PR (or auto-run via a git hook)	Fails the push if a commit touching artefacts/** is missing its provenance fields (data source / policy version / prior decision); previews which CODEOWNERS group will be required to review; runs tests if a test suite exists
scripts/install-hooks.sh	Once, per clone	Wires pre-pr-check.sh in as a pre-push git hook so it runs automatically

Setup:

1. Edit config/repo-governance.yml — replace repo.owner (<github-org-or-user>) with your real GitHub org/username, and adjust teams if your review-group names differ.
2. ./scripts/install-hooks.sh — makes pre-pr-check.sh run automatically before every push.
3. After your first git push, with the GitHub CLI installed and authenticated (gh auth login) with admin rights on the repo: ./scripts/bootstrap-repo.sh.

Both scripts were dry-run tested against a scratch repo (provenance pass/fail cases, CODEOWNERS path resolution, and a stubbed gh for the branch-protection payload) before being added here.