gtmvp.
00 · ARCHITECTURE · THE FIVE-LAYER SPEC SHEET

The five-layer spec sheet.
The receipts.

The marketing pages tell you GTMVP runs. This page tells you how. Five layers, every layer mapped to the file in the codebase that powers it. For the technical buyer who wants to verify the work before paying for it.

Run the diagnosticEngine room narrative
MODULES
22 per audit
SOLVERS
8 templates
SUBSYSTEMS
12 integrated
AGENTS
8 · through 2027
01 · OVERVIEW · LAYERS AT A GLANCE · 1 MIN READ

Five layers. One running engine.

Each layer below has a job. The operator decides. The modules write. The solvers prove. The subsystems integrate. The monitors watch. The layers stack — proof flows up.

Layer
Count
Purpose
L_05 · Operator
1 senior
Owns positioning calls, channel strategy, brand voice. Signs every brief before it ships.
L_04 · Modules
22 per audit
Five clusters: STRATEGY · COMPETITIVE · CONVERSION · CONTENT · TECHNICAL. Defined in lib/audit/generation/tools.ts.
L_03 · Solvers
8 templates
Z3 · MaxSAT · MiniZinc. Every recommendation is mathematically optimal under the founder's stated constraints.
L_02 · Subsystems
12 integrated
RivalScope, citation pack, prospect intel, ad library, stack detection, engine-agents pipeline, and more.
L_01 · Monitors
8 continuous
A_01–A_08 RivalScope agents. A_01 live today; the rest ship through 2027 on a published schedule.
02 · L_05 · OPERATOR LAYER · 1 MIN READ

One senior operator. On every account.

The agents propose. The operator decides. Every brief is read by a human before it leaves the engine. This is the layer that turns intelligence into a decision a CEO can defend in front of a board.

Senior GTM operator
Owns positioning calls, channel strategy, and offer architecture decisions. Signs every brief before it ships.
Brand reviewer
Sign-off on angles, voice, and message-market fit hypotheses before they ship.
Strategy lead
Quarterly review of competitive landscape, repositioning bets, channel mix.
03 · L_04 · MODULE LAYER · 2 MIN READ

22 modules per audit. Five strategic clusters.

Every audit produces 22 distinct deliverable modules, written by an orchestrated agent pipeline. The clusters run in dependency order — competitive intelligence feeds conversion strategy, which feeds the creative production plan, which feeds the technical implementation roadmap.

Cluster
Modules
Contents
CLUSTER_STRATEGY
3 modules
Executive Summary · ONE Mechanism · Implementation Priorities
CLUSTER_COMPETITIVE
3 modules
Customer Profiling + Archetypes · Competitive Positioning + Whitespace · Behavioral Intelligence + Channel Mix
CLUSTER_CONVERSION
6 modules
Pain Architecture · Decision Neuroscience · Belief Ladder · Offer Architecture · Awareness Matrix · MVMT
CLUSTER_CONTENT
6 modules
Creative Strategy + Production · Creative Angles · Ad Copy · Video Hooks · Video Scripts · Custom Creative Assets
CLUSTER_TECHNICAL
4 modules
Forensic Business Analysis · Macro Ad Account Review · Attribution + CRM Roadmap · AI & Automation Opportunity
SOURCE · lib/audit/generation/tools.ts → ALL_MODULES
04 · L_03 · SOLVER LAYER · 2 MIN READ

Eight solver templates. Provably optimal output.

Most AI marketing tools rank. GTMVP proves. Every recommendation that ships through the audit is the output of a constraint-solver run — Z3, MaxSAT, or MiniZinc — that returns the mathematically optimal answer under the founder's stated constraints. Or the unsat core, with the exact constraint that broke.

Template
Engine
What it proves
linear-allocation
Z3 Optimize
Provably optimal $$ split across 28 channels under budget, team size, and dependencies. Powers /channel-score.
max-min-distance
Z3 (Real)
Positioning whitespace · maximizes minimum distance to every rival in N-dim space. Powers /positioning-pass.
set-cover
Z3 (Bool/Int)
K-competitor selection with mathematical coverage of every feature dimension. Powers /competitor-map.
knapsack
Z3 (0/1)
SWOT priorities under founder weekly capacity. Picks the K with maximum impact-per-hour.
scheduling-with-deps
Z3 (Int)
TAM/SAM/SOM horizons with dependency DAG. Surfaces the critical path.
maxsat-claim-synthesis
PySAT RC2
Contradiction-free recommendation set. No two findings fight each other. Powers /gtm-audit synthesis.
quantifier-alternation
Z3 skolemized ∃∀
Competitive war-gaming · proves a move survives every competitor response, or returns the kill scenario. Powers /war-game.
assignment-with-diversity
MiniZinc
14-day editorial calendar diversity-optimized across pillars and platforms. Powers /content-calendar.
SOURCE · ~/.claude/plugins/marketplaces/gtmvp-gtm-agents/skills/solver-patterns/
05 · L_02 · SUBSYSTEM LAYER · 2 MIN READ

Twelve integrated subsystems. One engine.

GTMVP isn't an LLM wrapper. The engine integrates twelve services — each owning a slice of the GTM stack — into one operator workspace. Every audit pulls from every subsystem before any module writes.

RivalScope
Continuous competitor monitoring infrastructure. Workspace-scoped, classified events, weekly diff. Powers every A_0X agent.
Citation research pack
Pre-fetched Perplexity facts + URL snapshots before any module writes. Every claim is source-locked or it doesn't ship.
Prospect intel
ICP wizard + Apollo enrichment. 200 named decision-makers + intent signals per Battle Plan.
Ad library
Verified Meta + Google ad-creative archive. The competitor's actual ads — not screenshots of them.
Production ads
Operator-built ad variants generated against the audit findings. Ready-to-ship creative bundled with every Battle Plan.
Stack detection
Auto-detects the prospect's GTM stack on intake — ad platforms, attribution, CRM, CDPs, analytics.
Engine-agents pipeline
Kimi orchestrates the 22-module write loop with marker-deterministic boundaries and module-level state checkpoints.
Evidence + handoff
Every AuditEvidence row carries source URL, snapshot URL, retrieval timestamp, and source kind. Audit-grade citations.
Creative templates
Brand-locked templates for hooks, scripts, and ad copy. Operator extends, never starts from scratch.
Email sequences
Lifecycle automation for diagnostic delivery, upgrade, and onboarding. Resend-backed.
Billing (Stripe)
Subscription state machine for Diagnostic, Battle Plan, Pilot, Engine. Credits roll forward across tiers.
Calendly + Firebase
Operator hand-off scheduling + admin auth + per-customer workspaces.
SOURCE · lib/ (audit/{research, evidence, ad-library, prospect-intel, production-ads, engine-agents, generation}, rivalscope/, stack-detection/, creative-templates/, email/, billing/, calendly/, firebase/)
06 · L_01 · MONITOR LAYER · 2 MIN READ

Eight continuous agents. Shipping through 2027.

A_01 runs continuously today. The remaining seven ship on a published schedule — A_03 + A_07 in Q3 2026, A_06 + A_08 in Q4 2026, A_02 + A_04 + A_05 in Q1 2027. Engine subscribers get every agent the day it ships at no upcharge.

Agent
Cadence
What it watches
A_01 · Competitor recon
Continuous
Maps competitors and picks the K that mathematically cover every feature dimension. Weekly diff on pricing, hiring, releases.
A_02 · Product taxonomy
Q1 2027
Codifies feature/value/format gaps across the competitive set.
A_03 · Positioning
Q3 2026
Maps you + every rival as N-dim coordinates; solves for the position farthest from your nearest competitor. Provable whitespace.
A_04 · Offer design
Q1 2027
Synthesizes pricing, anchors, guarantees, bonus stacks.
A_05 · Angles
Q1 2027
Generates angles and proves the recommended set internally consistent — no two recommendations contradict each other.
A_06 · Channels
Q4 2026
Provably optimal $$ split across 28 channels under your budget, team size, and channel dependencies.
A_07 · Keywords + SERP
Q3 2026
Maps keyword universe, AI Overviews, share of voice movement.
A_08 · Trend pulse
Q4 2026
Hourly listening across X, LinkedIn, Reddit, podcasts, hiring, GitHub.
SOURCE · lib/rivalscope/types.ts → MonitorType + MONITOR_TYPE_TO_AGENT
07 · NEXT · 1 MIN READ

See the engine on your domain.

The architecture is the receipts. The Diagnostic is the proof. $129, 24 hours, 7-day money-back.

Run the diagnosticEngine room narrativeSee pricing