CLASSIFICATION: PUBLIC LICENCE: MIT DOC-ID: JGS-ARCHI-SKILLS REV 1.0.1

jgs-archi-skills

Agent-guided ArchiMate viewpoint creation in Archi. One orchestrator plus twelve specialists drive the existing JGS Archi Bridge MCP. Archi is the only canvas.

You govern scope. Skills consume MCP resources. They never copy ArchiMate tables and never patch the bridge.

§01 · What it is

A modelling run, not a diagram toy

You give the orchestrator a plain-language intent. It drafts a View Plan, grounds each viewpoint, and stops until you approve. Only then do layer specialists create or reuse elements in the live Archi model. Specialists are orchestrator-dispatched. MCP resources stay the ArchiMate reference.

§02 · Altitudes

Six layers, six questions

A technology architect already knows the stack. These rows say which ArchiMate layer answers which question, and which specialist the orchestrator may dispatch after you approve the View Plan. Legal relationships stay on the Bridge MCP.

LayerQuestion a technology architect asksWhat lands in ArchiSpecialist (dispatched)
MotivationWhy is this change on the board, and what outcome counts?Drivers, goals, principles, requirements, stakeholders, assessments, outcomesarchi-motivation
Strategy / capabilityWhat must the organisation be able to do, regardless of the current stack?Capabilities, resources, courses of action, value streamsarchi-capability-strategy
BusinessWho does the work, and which processes and services realize those capabilities?Actors, roles, processes, services, objects, eventsarchi-business
ApplicationWhich systems serve those services, and where is the overlap?Application components, services, interfaces, data objects, collaborationsarchi-application
Technology / physicalWhat hosts those systems, and what is as-is versus the landing zone?Nodes, devices, system software, networks, artifacts, physical equipmentarchi-technology-physical
Implementation / migrationWhat are the plateaus, gaps, and work packages, without redrawing the business?Work packages, deliverables, plateaus, gaps, migration roadmapsarchi-implementation-migration

Full layer treatment: the layer guide.

§03 · Agents

One pack, every main host

install.py already targets the main coding agents. Native folder copy for ZCode, Claude Code, OpenAI Codex, GitHub Copilot CLI, and OpenClaw. Gemini CLI as an extension. Cursor as project-local rules. Run python install.py --agent all for the user-global hosts, then --agent cursor in the repo that needs rules. Per-path table: docs/other-agents.md.

ZCode

Native folder. ~/.zcode/skills/<skill>/. --agent zcode

Claude Code

Native folder, namespaced. ~/.claude/skills/jgs/<skill>/. --agent claude

Cursor

Project-local .mdc rules. Run in the repo you want. --agent cursor

Gemini CLI

Extension under ~/.gemini/extensions/. --agent gemini

OpenAI Codex

Native folder. ~/.agents/skills/jgs/<skill>/. --agent codex

GitHub Copilot CLI

Native folder. ~/.copilot/skills/jgs/<skill>/. --agent copilot

§04 · How you ask

Plain language. One entrypoint.

01

Capability map

Finance and ops agree which capabilities run quote to cash, and which applications already cover them.

Altitudes: strategy / capability, business, application.

/archi-orchestrator invoice-to-cash capability map for finance and ops

02

Merger current vs target

Two onboarding journeys, a dual-run plateau, work packages that retire a core without a silent cutover.

Altitudes: motivation, business, application, implementation / migration.

/archi-orchestrator Northridge and Vale insurance merger: current vs target customer onboarding, with a migration roadmap

03

NL change / impact

A second CRM on an existing model, impact visible, shared customer data not duplicated.

Altitudes: application, business (impact only).

/archi-orchestrator on the current model, add a second CRM for the European branch and show impact; do not duplicate shared customer data

04

Motivation-only board pack

Drivers, goals, and outcomes for quote time. No applications on the canvas.

Altitudes: motivation only.

/archi-orchestrator board pack: drivers, goals, and outcomes for cutting quote time from 4 hours to 30 minutes. No applications.

05

Technology landing zone

On-prem TMS to a landing zone. Plateaus and work packages only. Business stays as it is.

Altitudes: technology / physical, implementation / migration.

/archi-orchestrator move the legacy TMS to a cloud landing zone; plateaus and work packages only. Do not redesign the business.

06

Layout and QA only

An existing view is unreadable. Relayout and report compliance. Do not add elements.

Altitudes: none (layout and model-qa only).

/archi-orchestrator the Application Support view is unreadable. Relayout it and report compliance issues. Do not add elements.

Named deliverable

You named the views or layers. The run stops when they exist and the out-of-scope holds.

Outcome until

You named what must be true for the people who care. The run keeps going, in the layers you scoped, until those checks are visible on the model.

ArchiMate Goal is a motivation element. The run stopping is a stop rule, not a Goal. Approving the View Plan signs that stop.

/archi-orchestrator is the only modelling entrypoint. /jgs-upstream-feedback is the pack-defect door. Do not invoke layer specialists yourself.

§05 · Why

A shared pack that gets better in public

Use this when you want ArchiMate in Archi, not a slide deck the model cannot replay. You approve the View Plan. Specialists write only the layers you named. The language reference stays on the Bridge MCP.

The other reason is the loop. When a run hits a pack or MCP-contract hole that any user would hit, the orchestrator can offer to file it, or you invoke /jgs-upstream-feedback. That issue lands on GitHub. Fixes ship in the next pack. Everyone who installed the skills gets the same improvement. Local architecture stays local. Only agnostic defects go upstream.

Questions, worked runs, and "has anyone modelled this" belong on GitHub Discussions. Defects that should change the pack stay on Issues. There is no separate Discord yet. Discussions is the public room.

§06 · Eval

Frozen scenario, dedicated model

Maintainer loop on the frozen Meridian Freight scenario in a dedicated eval model (JGS Eval Loop). This is not a public skill-versus-no-skill A/B. Iter-0 metamodel findings of 40 were largely an allowlist gap on an already bridge-legal model. The instrument and CREATE_PATH rules tightened. Treat the loop as a lower bound on pack quality, not a marketed score.

Frozen Meridian Freight scenario on a dedicated eval model (JGS Eval Loop). Maintainer loop, not a public skill-versus-no-skill A/B. Eval folders stay unpublished.
DimensionIter-0Iter-1
Metamodel fidelity (compliance_validate findings)400
Naming conflicts00
Documentation coverage findings0 (after backfill)0 (first pass)
Layout check findings0 (after live discovery)0 (deterministic)
Rationale bundle00
Motivation-to-technology chaincompletecomplete
§07 · Examples

Three frozen jobs

Prompt cards on GitHub. Live runs still stop at the View Plan gate. A finished mill walk (Hatherley Plate, author-run) is on hatherley.html; the working models and paste fences live in jgs-archi-skills-we.

Invoice to cash

/archi-orchestrator invoice-to-cash capability map for finance and ops

Views: Capability Map, Order-to-Cash Operations, Application Support.

Pass: a path from a capability through a process to an application.

p01-invoice-to-cash.md

Insurance merger

/archi-orchestrator Northridge and Vale insurance merger: current vs target customer onboarding, with a migration roadmap

Views: Motivation Overview, Current and Target Customer Onboarding, Application Landscape, Migration Roadmap.

Pass: plateaus or work packages exist in the model; Northridge and Vale systems stay distinct on the current view.

p02-insurance-merger.md

TMS landing zone

/archi-orchestrator move the legacy TMS to a cloud landing zone; plateaus and work packages only. Do not redesign the business.

Views: Technology Current, Technology Target, Migration Roadmap.

Pass: no new business process invented to justify the move.

p08-tms-cloud-migration.md

§08 · Install

Clone, install, restart, ask

Prerequisites: Archi with the JGS Archi Bridge plugin, MCP at http://127.0.0.1:18090/mcp, Python 3.10+ (stdlib only). Install into every user-global host, or pick one. Cursor is project-local, so run it in the repo that needs rules. Paths: docs/other-agents.md. The long agent-install prompt stays in the README.

# Python 3.10+, stdlib only. Archi + JGS Archi Bridge on loopback.
git clone https://github.com/jgsystemsconsulting/jgs-archi-skills
cd jgs-archi-skills
python install.py --agent all     # ZCode, Claude Code, Codex, Copilot, OpenClaw, Gemini
python install.py --agent cursor  # project-local .cursor/rules
python install.py --list-agents
python install.py --dry-run

Restart the agent so it discovers the skills. Then call the orchestrator. It stops at a confirmation gate. Approve the View Plan before anything is written into Archi.

# technology-architect pair; full catalog: docs/prompts/README.md
/archi-orchestrator invoice-to-cash capability map for finance and ops
/archi-orchestrator move the legacy TMS to a cloud landing zone; plateaus and work packages only. Do not redesign the business.

Usage: docs/skill-usage.md. Starter catalog: docs/prompts/README.md.