Anatomy of the
GTM agent system.
This is the operating layer behind the demos—not a gallery of generic AI personas. Each agent has a narrow responsibility, a reusable SKILL.md, a read-only default, and a defined output contract.
AGENT / SKILL MANIFEST
⌄every recommendation has an owner · every agent has a boundary
Workflow routing
choose the smallest relevant specialist setA request is matched to the most relevant specialist. When several are needed, their outputs are combined into one reviewable recommendation.
routing logic.tomlAn agent setup file that defines its name, purpose, permissions, and operating boundaries.SKILL.mdAn instruction guide that defines how the specialist performs its work and structures the output..codex/agents/
01growth-command-analyst.tomlDiagnoses funnel, channel, attribution, pipeline, CAC, payback, and anomalies. ready⌄
.agents/skills/gtm-growth-command/SKILL.md$gtm-growth-commandChannel, funnel, attribution, pipeline, CAC and payback evidence.
Executive decision brief, next test, guardrails and measurement plan.
Reconciles units and attribution windows; never executes budget changes.
02revenue-prioritization-agent.tomlScores and routes accounts using separate, explainable fit and intent evidence. ready⌄
.agents/skills/gtm-revenue-agent/SKILL.md$gtm-revenue-agentAccount records, fit evidence, intent signals and routing rules.
Prioritized accounts, routing rationale and reviewable outreach drafts.
No sensitive-trait inference, CRM update, or message send without approval.
03market-radar-researcher.tomlTurns competitor changes into evidence-linked commercial alerts and response briefs. ready⌄
.agents/skills/gtm-market-radar/SKILL.md$gtm-market-radarCurrent and prior pricing, product, positioning, advertising, review, or release evidence.
Prioritized change alert with evidence IDs and proportionate response options.
Separates observed change from inference; never publishes competitive claims.
04customer-signal-strategist.tomlSynthesizes customer evidence into segment messages, objections, and bilingual variants. ready⌄
.agents/skills/gtm-customer-signal/SKILL.md$gtm-customer-signalInterviews, reviews, support tickets, feedback, and win/loss notes.
Segment insight map, objections, positioning, and English/Chinese variants.
Preserves source meaning; never invents quotes or false consensus.
05experiment-design-agent.tomlConverts growth problems into ranked, measurable experiments and decision rules. ready⌄
.agents/skills/gtm-experiment-studio/SKILL.md$gtm-experiment-studioGrowth problem, audience, intervention, metrics, constraints, and instrumentation.
Prioritized hypothesis, experiment design, guardrails, and readout plan.
Defines success and stop rules before launch; never claims unobserved results.
06capital-allocation-agent.tomlModels channel allocation with marginal returns, constraints, ranges, and staged moves. ready⌄
.agents/skills/gtm-capital-optimizer/SKILL.md$gtm-capital-optimizerBudget, marginal returns, capacity, saturation, forecast ranges, and constraints.
Constraint-aware investment mix with staged reallocations and downside limits.
Checks budget conservation; never moves money or alters campaigns.
07experience-personalization-agent.tomlSelects explainable website experiences from approved, non-sensitive signals. ready⌄
.agents/skills/gtm-experience-ai/SKILL.md$gtm-experience-aiAllowed firmographic and behavioral signals, content rules, and proof points.
Selected message, proof, CTA, fallback, and controlled experiment.
No sensitive traits; requires confidence threshold, fallback, and control group.
08market-launch-strategist.tomlRanks adjacent markets and creates a risk-aware 90-day validation plan. ready⌄
.agents/skills/gtm-market-launcher/SKILL.md$gtm-market-launcherOpportunity, product fit, access, economics, execution risk, and weights.
Entry thesis, ICP, positioning, channels, risks, and validation roadmap.
Separates evidence from assumptions and gates scale behind validation.
09journey-orchestration-agent.tomlDesigns behavior-triggered lifecycle journeys with suppression and escalation. ready⌄
.agents/skills/gtm-journey-agent/SKILL.md$gtm-journey-agentLifecycle stage, behaviors, account evidence, channels, and eligibility rules.
Journey steps, channel drafts, suppression, frequency, stop, and escalation rules.
Never contacts customers or changes a lifecycle system without approval.
10creative-strategy-agent.tomlDevelops evidence-grounded creative territories and clean performance tests. ready⌄
.agents/skills/gtm-creative-studio/SKILL.md$gtm-creative-studioCampaign goal, audience, offer, customer evidence, brand rules, and results.
Distinct creative territories, claims, formats, and disciplined test cells.
Traces ideas to evidence; never publishes or launches creative.
11ai-visibility-strategist.tomlAudits answer-engine visibility and builds evidence-led GEO and citation plans. ready⌄
.agents/skills/gtm-ai-visibility/SKILL.md$gtm-ai-visibilityAI-answer samples, prompt lineage, citations, content, community, and review evidence.
Visibility audit and prioritized GEO, citation, community, and review plan.
No fake reviews, sockpuppets, deceptive seeding, or guaranteed rankings.
demo/shared fixtures, contracts, and regression criteria
governeddecision-packet.schema.jsonfacts · assumptions · recommendation · confidence · approvals
evals.jsonhonesty · math discipline · evidence handling · human control
fixtures.jsonreproducible fictional demo records
From question to
reviewable recommendation.
Routing logic selects only the specialists needed. Evidence stays attached to the recommendation, and any external action stops for human approval.