GTM Engineering skills
GTM Engineering Skills-md for Claude, Codex & AI Agents
GTM engineering skills are SKILL.md files that help Claude, Codex and other AI agents build and run go-to-market systems: Clay tables, enrichment waterfalls, ICP scoring, signal-based outbound, AI-written personalisation, CRM syncs and the QA that stops quiet failures reaching prospects. The first live skill checks every row of a GTM table against a written spec before anything is sent or synced.
13 of 13 skills written, 3 with full pages, version 1.0.0, October 2026.
GTM Engineering Audit
Audit a GTM engineering stack end to end: ICP, providers, Clay tables and workflows, enrichment, scoring, signals, AI copy, data quality, CRM and sequencer sync, sending infrastructure, QA, costs, documentation and reporting, with a funnel from source to revenue, unit costs, a 0 to 3 scorecard, a 90-day plan and governance rules.
Clay Table Design
Design Clay and similar GTM tables that are cheap, fast and debuggable: a table spec, one job per table, staged columns from input and cheap qualification to paid enrichment, scoring, signals, AI copy, QA and output, run conditions on every paid column, naming conventions, dependency order, sample tests and reusable templates.
Enrichment Waterfalls
Design enrichment waterfalls for emails, phones, contacts and company data: success definitions, provider benchmarks by region, ordering by cost per valid result, stop-on-success run conditions, email verification with catch-all policies, confidence fields, regional variants and monthly monitoring of match, bounce and cost.
ICP and Account Scoring
Score accounts and contacts for GTM workflows: hard exclusions, a fit score from won and lost deal data, a timing score from expiring signals, contact scores by role and seniority, a fit-by-timing tier matrix with actions, validation against past deals and monthly monitoring.
Signal-Based Outbound
Run outbound on buying signals: signals chosen from past wins, precise definitions and sources, strength and freshness decay, a play library matching signals to people, angles and calls to action, detection workflows, relevant non-creepy openers, routing within 48 hours and measurement by signal.
AI Copy Columns
Design AI copy columns for outbound in Clay or similar tools: one job per column, clean short inputs, prompts with rules, examples and fallbacks, gating to qualified tiers, a five-part scoring rubric, QA for placeholders and invented facts, prompt rotation and reply-rate monitoring by version.
Data Deduplication and Normalisation
Clean GTM data: normalisation rules for domains, company names, people, emails, countries, phones, titles, seniority, departments and industries, company matching on root domain, contact matching on email and LinkedIn, match confidence, survivor rules, CRM lookups before creating records and duplicate prevention.
CRM Sync and Routing
Sync GTM tables to CRMs and sequencers safely: field mapping with direction and overwrite rules, lookups before create, protected rep and system fields, ownership and routing by territory, tier and capacity, enrolment and suppression rules, source stamping, error handling and weekly monitoring.
Outbound Infrastructure and Deliverability
Set up and run cold outbound infrastructure: secondary domains, mailboxes, SPF, DKIM and DMARC, warm-up, capacity planning and moderate per-mailbox limits, verified lists, deliverable plain-text copy, Google and Yahoo sender requirements, health thresholds for bounces and complaints and an incident response plan.
GTM Workflow QA
QA every row of a Clay or other GTM table against a written spec before sending or syncing: required fields, formats, ICP and exclusion rules, duplicates, CRM mapping, copy checks for placeholders, length, banned phrases, invented facts and tone, code checks plus batched AI judgement, errors grouped by cause and a go or no-go report.
GTM Cost Control
Control GTM engineering costs: a cost ledger by tool, provider and workflow, credit attribution by table and column, unit costs from enriched contact to meeting and opportunity, a waste checklist, caching and run conditions, budgets with alerts, a target cost per meeting from deal economics and scale, fix or stop decisions.
GTM Workflow Documentation
Document GTM engineering workflows so others can run and debug them: workflow specs, data dictionaries with meaning, source, run condition and cost, flow diagrams, runbooks for common failures, access lists without secrets, change logs, handovers and quarterly reviews.
GTM Reporting
Report GTM and outbound results from source to revenue: agreed funnel definitions, attribution by workflow, segment, signal, play and copy version, reply classification, health and unit-cost metrics, dashboards for operators, teams and leaders, weekly and monthly templates and experiments on the weakest stage.
No skill matches that search. Clear it or choose All.
What GTM engineering skills do for an AI agent
GTM engineering turns go-to-market ideas into working systems. A typical workflow finds companies, enriches contacts through several data providers, scores fit, writes personalised lines with AI and pushes records to a CRM or sequencer. Each step can break quietly: a provider returns nothing for one country, a formula fails on unusual data, an AI column leaves a placeholder or invents a fact, and hundreds of messages go out before anyone notices.
These skills give an agent such as Claude Code a disciplined way to build and check that work. They start from a written spec, separate deterministic checks that code can do exactly from judgement calls where AI helps, group failures by the step that caused them and stop launches until results meet agreed thresholds. The set will cover table design, waterfalls, scoring, signals, AI copy, deduplication, CRM sync, infrastructure, cost control, documentation and reporting.
Clay, enrichment, outbound and QA
The skills work with Clay and similar tools, spreadsheets and CRMs. They sit beside the CRM and Lifecycle, Product Marketing and Content skills, which cover ICPs, data hygiene and copywriting in more depth.
Where to start: QA before you launch
Start with gtm-workflow-qa. Write a short spec for your table: required fields, formats, ICP and exclusion rules, duplicate limits, copy rules for AI-written lines, CRM field mapping and the fail rate you will accept. Export the table as CSV and let the agent check every row, using a Python script for exact rules and AI judgement only for tone, relevance and invented facts.
The report groups failures by cause, such as a provider step skipped for one country or an AI column that ran before first names were filled, suggests fixes and gives a go or no-go. Fix the workflow, re-run QA and launch only when it passes. Most teams find problems in their first full QA that spot checks never showed.
Run the same QA every time you change a step, add a provider or reuse a table template, and schedule it daily for always-on workflows so new rows are checked before they reach your sequencer or CRM.
As the set grows, the order will follow how GTM systems are built: audit what exists, design clean tables, set enrichment and scoring rules, add signals and AI copy, clean and sync the data, protect deliverability, control costs and document everything so the next person can run it.
Each new skill will include its own spec templates and checks so the whole set works together.
Teams new to GTM engineering can start with QA alone and add the other skills as their workflows grow.
The 13 skills in this set, and the rules behind them
Thirteen skills are planned for this set. All 13 are live with full pages.
| Skill | What it does | Status |
|---|---|---|
gtm-engineering-audit | Audit your GTM data, tools, workflows and costs | Live |
clay-table-design | Clay tables that are easy to run, debug and reuse | Live |
enrichment-waterfalls | Provider waterfalls with coverage, cost and accuracy rules | Live |
icp-account-scoring | Scoring accounts and contacts against your ICP | Live |
signal-based-outbound | Triggers from hiring, funding, tech and intent signals | Live |
ai-copy-columns | AI-written personalisation that stays specific and true | Live |
data-deduplication-normalisation | Clean, deduplicated and standardised records | Live |
crm-sync-routing | Syncing to CRMs and sequencers with owners and rules | Live |
outbound-infrastructure | Domains, mailboxes and sending limits for deliverability | Live |
gtm-workflow-qa | Row-by-row QA against a written spec before launch | Live |
gtm-cost-control | Credits, provider costs and cost per qualified lead | Live |
gtm-workflow-documentation | Specs, runbooks and handovers for GTM systems | Live |
gtm-reporting | Pipeline, reply and conversion reporting for GTM workflows | Live |
The rules behind these skills
| Rule | Why it matters |
|---|---|
| Write the spec first | Nobody can QA or debug a workflow without agreed requirements |
| Check every row | Quiet failures hide in small percentages of large tables |
| Code for rules, AI for judgement | Deterministic checks are cheaper and exact; AI handles tone and facts |
| Fix the step, not the row | Errors grouped by cause show which provider, formula or prompt broke |
| Nothing ships on a fail | Sends and CRM syncs wait for a go decision |
| Count the cost | Credits and provider fees add up across waterfalls |
Example prompts
Each skill triggers on everyday wording. These prompts are good starting points:
- QA: “QA this Clay export against our spec before we launch.”
- Spec: “Write a QA spec for our fintech outbound table.”
- Copy checks: “Check every AI opener for placeholders and invented facts.”
- Causes: “Group these failures by the workflow step that broke.”
- Schedule: “Run QA daily on new rows and alert me on failures.”
- CRM: “Check these records are ready to sync to HubSpot.”
What the agent needs from you
GTM QA and builds depend on a few inputs:
- A written spec: fields, formats, rules, copy limits and thresholds.
- The data: a CSV export or API access to the table.
- Exclusion lists: customers, competitors and do-not-contact lists.
- CRM field names: for mapping checks before sync.
- A runtime: Claude Code or another agent that can run Python with pandas.
If the spec is missing, the skill drafts one from your brief and workflow for you to approve first.
What an agent cannot do, and myths to ignore
An agent can check, report and suggest fixes, but it cannot decide your targeting, guarantee data provider accuracy or confirm that outreach meets data protection and email marketing laws in each country. Legal and compliance review stays with your team.
| Myth | What works better |
|---|---|
| Spot-checking ten rows is enough | Check every row; small error rates hide in large tables |
| AI should check everything | Code checks exact rules; AI judges tone and facts |
| Fix bad rows by hand | Fix the step that produced them, then re-run |
| More providers always help | Waterfalls need coverage, cost and accuracy rules |
| Personalisation means mentioning anything | Lines must use real, relevant data from the row |
Frequently asked questions
What is GTM engineering?
Building the data, automation and AI workflows behind go-to-market work, such as enrichment, scoring, personalised outbound and CRM syncs.
How do I QA a Clay table?
Write a spec, export the table and check every row with code for exact rules and AI for tone and facts, then fix causes and re-run.
What goes in a QA spec?
Required fields, formats, allowed values, ICP and exclusion rules, duplicates, copy rules, CRM mapping and fail thresholds.
Can Claude Code run the QA?
Yes. It runs a Python checker on the CSV and batches AI checks for copy, then writes the report.
How do I stop AI copy inventing facts?
Only allow lines to use signals present in the row and check every line against that data before sending.
Are these skills free?
Yes, under CC BY 4.0, as long as you credit Afeez Adebayo and link to the source page.
Sources and more skills
Checked on 13 October 2026. Related sets: CRM and Lifecycle Marketing, Product Marketing and GTM and Growth Marketing.
