Protocol — approved, execution not started

First-Run Protocol: Vertical → Certified Snapshot Template

The route from “Commercial Electrical is the industry” to a certified, versioned GoHighLevel snapshot template — a frozen backup location, a working clone, and a proven-installable snapshot. Eight gates, seven stages, working backwards from the endpoint.

First run Commercial Electrical Contractors Endpoint commercial-electrical-v0.1.0, Gate 7 Source FIRST-RUN-PROTOCOL-vertical-to-snapshot-template.md

Model routing Fable Opus Sonnet Human ⚑ External research Fable — gate adjudication, the token contract, red-team, release sign-off: the few moments a wrong call poisons everything downstream. Opus — Storm orchestration, journeys, all customer-facing copy: judgment work with volume. Sonnet — schemas, scripts, lexicon assembly, QA, citation verification: mechanical work, however large. Research and verification are not cost-constrained — the savings come from routing volume down, never from thinning evidence.

The four locations

Defined now, built in Stage 6–7. Every release moves through this sequence — construction is disposable, the template is sacred.

CE-BUILD

Release candidate

Messy is fine. Test data lives and dies here. Rebuilt per release.

→

CE-TEMPLATE

Certified master · the backup

Frozen. No experiments, no client data, no manual edits — ever.

→

CE-SOURCE

Clone of Template

The working snapshot source. Future changes land here, get re-certified, re-promoted.

→

CE-QA-*

Throwaway install target

Snapshot pushed in, full QA run, then deleted. Never reused.

Every version: CE-BUILD passes Gate 6 → snapshot cut → pushed to fresh CE-QA (override:false) → QA passes → CE-BUILD promoted: TEMPLATE refreshed, SOURCE re-cloned → version tagged

Two-level Storm research

The macro layer builds the template once. The micro layer localizes it to every customer, at fulfillment time — this is what lets one spine wear many sub-niche faces.

Vertical Storm

Once · Stage 1 · full 5-lens pipeline

Practitioner, Academic, Skeptic, Economist, Historian + contradiction map + verification
Builds: operating model, lexicon, ICP, sub-niche map, geographic matrix
Feeds the template itself — every pipeline stage, field, and AI boundary traces back to it

Sub-niche Storm — W-C09

Per customer · onboarding · scoped 3-lens run

Practitioner, Economist, Skeptic + verification — localizes the macro layer, doesn't rebuild it
Inputs: client's sub-niche, geography, service mix, intake facts
Outputs: KB candidate facts (flagged for client confirmation, never auto-ingested), copy emphasis, the $47 Lead Gen Strategy, local campaign intel

The seven stages

Every stage supplies the input the next one requires. A gate blocks progress until its condition is met — no stage begins on faith.

Context lock, workspace & platform facts

Stage 0

A clean workspace and every GHL platform unknown answered before anything depends on it — the interpolation matrix moves here so token-first authoring in Stage 6 isn't built on hope.

0.1

Context lock

Settled decisions, System A/B/C boundaries, forbidden patterns, contradictions between governing docs

Sonnet

0.2

Workspace init

Sonnet

0.3

Decision register

Sonnet

0.4

GHL platform fact-finding

Custom-values interpolation matrix (field by field) · KB create→write→assign→retrieve · Agent DTO round-trip · snapshot push round-trip

Sonnet⚑

0.5

Location provisioning

Create CE-BUILD + first CE-QA shell

Human

Artifactsdocs/context-lock.md · decisions/decision-register.md · recorded platform-test results

GATE 0Workspace exists, context lock written, all four platform tests answered. No research proceeds on unverified platform assumptions.

Vertical intelligence

Stage 1

The evidence-backed operating model, lexicon, sub-niche map, and ICP the whole template stands on.

1.1

Vertical Storm run

Full 5-lens pipeline — demand-side, industry structure, practitioner reality, the bid chain, adjacent-software gap, geographic signal

OpusSonnet⚑

1.2

Evidence register

Sonnet

1.3

Vertical operating model

The single most important artifact — every downstream stage/field/boundary traces here

OpusFable

1.4

Geographic priority matrix

Sonnet

1.5

Lexicon

Every term must trace to a practitioner source — no invented industry-speak

SonnetOpus

1.6

Sub-niche candidate map

The bridge-page layer — 4–6 candidates, what varies vs. what stays structural

Opus

1.7

ICP + exclusions

OpusFable

Artifactsresearch/commercial-electrical-infrastructure-briefing.html · evidence-register.md · vertical-operating-model.md · geographic-priority-matrix.md · lexicon/commercial-electrical.yaml · sub-niches/sub-niche-map.md

GATE 1Product thesis approval — hard stop for your sign-off. Decision packet: segment, size band, geographies, buyer, workflows, disqualifiers, evidence, red-team, sub-niche map. Go / revise / stop.

Product lock + sub-niche Storm protocol

Stage 2

One product, one sentence, one primary workflow — and the per-customer research machine specified.

2.1

Product charter

Scope wall verbatim: not dispatch, not job costing, not FSM, not PM, not ERP, not estimating software, not ad management

Fable

2.2

Operating lanes

Primary: inquiry→qualify→site-walk→follow-up. Secondary: reactivation

Fable

2.3

Sub-niche Storm protocol

The reusable per-customer run — trigger, inputs, 3-lens scope, outputs, cost target, registered as W-C09

OpusFableSonnet

2.4

Sub-niche variation matrix

Token / KB / copy / agent-fact variation per sub-niche — structural proposals rejected here

Opus

Artifactssnapshot-spec/product-charter.md · storm-protocol/subniche-storm-protocol.md · sub-niches/variation-matrix.md

GATE 2Product explainable in one sentence, mapped to one primary workflow; sub-niche protocol specified with a cost target.

The contractor journey

Stage 3

The full operational path a project opportunity travels — specified before any form, pipeline, or workflow exists.

3.1

Real-artifact sampling

12–15 real contractor estimate-request forms — grounds intake in reality, not CRM convention

Sonnet⚑

3.2

Project-opportunity journey

Every step: entry, page, fields, qualification rule, CRM write, pipeline movement, exception path. AI collects, never technically approves

OpusFable

3.3

Reactivation journey

Opus

Artifactsresearch/intake-artifact-sample.md · customer-journeys/project-opportunity-journey.md · reactivation-journey.md

GATE 3Journeys approved before a single form, pipeline, or workflow is created.

Snapshot asset specification

Stage 4

Every V1 asset named, owned, dependency-mapped, and QA-checked — on paper, before GHL is touched.

4.1

Asset registry

6 pages, 6 forms, 2–3 calendars, 14-stage pipeline, 20-workflow catalog, 2 AI agents, 4 KB containers

Sonnet

4.2

Workflow catalog validation

CE-W01…W20 — a workflow with no evidence behind it goes to backlog, not build

Sonnet

4.3

Pipeline map

Stage names are object names → API renames, never tokens

Sonnet

4.4

Form-field map

Sonnet

4.5

AI agent spec

May-do / must-not-do against verified Agent DTO fields. SMS excluded at creation; mode:off at creation

Opus

4.6

KB taxonomy

4 coarse containers, not one KB per document

Sonnet

Artifactssnapshot-spec/asset-registry.yaml · workflow-catalog.md · pipeline-map.md · ai-agent-spec.md · knowledge-base/kb-content-schema.md

GATE 4No GHL construction until every V1 asset has owner, dependencies, personalization method, and QA check.

Manifests & the token contract

Stage 5

The snapshot generic, the vertical language fixed, the client facts replaceable — provably. The keystone stage.

5.1

Vertical manifest

Identity, taxonomy, pipeline stages, labels, agent template, sub-niche variant blocks

Sonnet

5.2

Token dictionary

Filtered by the Stage 0 interpolation matrix. Missing required token = compile failure. No silent blanks

SonnetFable

5.3

Client build schema

Includes sub_niche + storm_run_id — the W-C09 linkage

Sonnet

5.4

Paper personalization test

One fictional contractor: every value resolves, zero workflow-body edits, or the gate fails

SonnetFable

Artifactsmanifests/vertical-01.yaml · token-dictionary.yaml · client-build.schema.json

GATE 5The paper contractor personalizes completely through manifests alone — no exceptions.

Build the release candidate

Stage 6

CE-BUILD becomes a complete, working, token-first vertical.

Reset CE-BUILD

Sonnet

API-supported assets

Fields, tags, pipeline, calendars, custom-values scaffold — scripts reused every release as W-F05

Sonnet

UI-only assets

Funnel pages, forms, workflow builder — Factory-tier, slow and supervised is fine

Human

Copy authoring

Every message token-first — you approve tone on a sample, patterns propagate

Opus

AI agents

Created via API at mode:off, no SMS channel

Sonnet

KB containers + placeholder content

Sonnet

Dashboards + attribution

Human

Local QA

Test contacts through every journey — failures logged as build defects

Sonnet

Token leak scan

Zero unresolved {{...}} — becomes W-Q02

Sonnet

Test-data teardown

Becomes W-Q08

Sonnet

Artifactsbuild-sequence.md · dependency-graph.yaml · form-field-map.md (as-built) · pipeline-map.md (as-built) · qa/release-qa-plan.md

GATE 6Full contractor journey completes with test records, passes the token leak scan, is clean of test data.

Snapshot release, template & clone

Stage 7 · endpoint

The versioned artifact, the frozen backup, and the working source.

Create commercial-electrical-v0.1.0

Push to fresh CE-QA

override: false, always

Sonnet

Full install QA

Asset inventory, journey smoke test, token resolution, KB live-retrieval test, agent off→auto-pilot→off round-trip

SonnetOpus

Snapshot mutation matrix

One-time insurance — what a controlled push preserves vs. overwrites

Sonnet

Promote

CE-BUILD certified → create CE-TEMPLATE (frozen) → clone to CE-SOURCE → delete QA location

Release artifacts

Release notes, manifest with asset inventory + token dictionary hash

SonnetFable

Artifactsqa/snapshot-mutation-matrix.md · release/release-notes.md · commercial-electrical-v0.1.0-manifest.json

GATE 7Protocol complete. A blank location receives the snapshot with override:false and reproduces the full inventory and journey. Template frozen, source cloned, version tagged.

What happens after — separate plans

Follow-on	Governed by	Consumes from this protocol
Activation spine	Fulfillment map, Phase 2	Snapshot, token dictionary, client-build schema, W-C09 protocol
Tier1 acquisition funnel	Directive Phase 9 + funnel-as-a-service flow	Sub-niche map, lexicon claims, geographic matrix
Pilot — 3–5 contractors	Directive Phase 10	Everything; variations classified vertical_standard vs. client_*

First-Run Protocol — Tier1 Marketing · visual companion to FIRST-RUN-PROTOCOL-vertical-to-snapshot-template.md