GTM skills & agents

GTM skills that upgrade how your firm sells

A library of my GTM frameworks and skills. It will cover the whole client journey: strategy, marketing, lead generation, sales, account management.

start herenpx skills add victor-shulga/max
strategynpx skills add victor-shulga/gtm-strategy-skills
marketingnpx skills add victor-shulga/marketing-engine-skills
outboundnpx skills add victor-shulga/outbound-engine-skills
salesnpx skills add victor-shulga/sales-engine-skills
contentnpx skills add victor-shulga/content-engine-skills
single agentsnpx skills add victor-shulga/gtm-skills
retentionnpx skills add victor-shulga/account-management-skills
infrastructurenpx skills add victor-shulga/mcp-skills

Requires Skills CLI — or grab the ZIP from any skill's page, no terminal needed

How to use the skills →
Skills are the parts. The system is how they connectThe nine builds this library assembles into: what runs after what, and the number each block produces.
See the stack

Master Skills

4 flows

Multi-step systems — each one orchestrates a dozen sub-skills into a finished result.

Bundle24 skills

The Outbound Engine

The outbound block as a conveyor: signal catalog → job-posting sourcing → scored list → verified contacts → sequence → replies → reports. A master skill runs the path and holds the gate between steps.

View →
Master Skill13 sub-skills

The GTM Strategist

A 13-step cold-audit that turns an agency's URL into a full go-to-market strategy and a 30-90-180 action plan.

View →
Master Skill8 sub-skills

The Content Engine

An 8-step LinkedIn content system: strategy → research → weekly plan → production → engagement.

View →
Kit2 skills

LinkedIn Strategy Kit

The foundation layer: build a LinkedIn strategy config, then audit the profile against it. The first two steps of the Content Engine.

View →

Agents

17 skills

Single skills — each does one job well. You type /name — and Claude knows the job.

/hypo-generatorAgent

Hypothesis Generator

From an agency site → a weighted ICP × signal × offer hypothesis matrix.

View →
/reply-auditAgent

Reply Audit

Forensic audit of outbound replies → root cause (targeting vs pitch) → 9-section report.

View →
/prospect-scoringAgent

Prospect Scoring

Deterministic 100-pt ICP rubric with anti-ICP gates, before SDR outreach.

View →
/offer-factoryAgent

Offer Factory

From a site → 5–10 testable offer-bets (ICP × pain × mechanism), scored.

View →
/agency-signal-sourcerAgent

Signal Sourcer

Buying-signal engine: detect, decay windows, scoring, heat tiers, signal-to-action plays.

View →
/weekly-outreach-reportAgent

Weekly Outreach Report

Narrative weekly outreach report (7+2 sections) with data-quality warnings first.

View →
/reply-objection-handlerAgent

Objection Handler

One reply → classify → one ready-to-send message + a library of proven templates.

View →
/meeting-prepAgent

Meeting Prep

Pre-call brief ≤500 words for a booked discovery or demo call.

View →
/proposal-generatorAgent

Proposal Generator

Two proposals (call-deck + send version) on the client's brand, with a critique checklist.

View →
/anticopywriting-aiAgent

Anti-Copywriting AI

Strips the AI tone from UK/EN text — clichés, filler, over-formatting, self-importance.

View →
/sequence-writerAgent

Multichannel Sequence

An email and LinkedIn sequence for one hypothesis: the anchor first, then capacity, and only then the writing. Never asks for a call.

Look inside →outbound
/page-builderAgent

Page Builder

A pipeline for website pages: a Kit per page type, an architecture gate before the first word, six passes, a blind score of 85/100.

Look inside →page
/watchAgent

Watch

Claude watches videos and call recordings: on-screen frames, a local transcript, speakers and objections with timecodes.

Look inside →video
/design-system-generatorAgent

Design System Generator

A client's URL → a single-file HTML design system built on their real colors and fonts.

View →
/offer-ladderAgent

Offer Ladder

Free → low → mid → high tiers around one core transformation, checked for cannibalization.

View →
/case-study-writerAgent

Case Study Writer

A finished project → a 12-section case study, result-first, plus a one-pager and a designer brief.

View →
/prospect-profilerAgent

Prospect Profiler

A scored account list → one 60-second pre-touch card per account, before anyone writes copy.

View →
The whole library costs almost nothing in tokens.

Claude reads only each skill's short header until that skill is actually needed.

FAQ

What is a skill?
A skill is a packaged set of instructions, code and resources that hands Claude ready expertise for one specific job. A bundle of experience, loaded on demand. Install it once, then call it with a single command /name.
How do I install one and get going?
Three routes, written out step by step on every skill's page. With a terminal — Skills CLI: the command takes one set at a time, which is why there are five at the top of this page, one per repository. A single skill has its own command on its own page. Without a terminal — the Download the ZIP button on the skill's page: drop the folder into ~/.claude/skills/, or upload the archive to Claude Desktop under Settings → Capabilities → Skills. The third route is taking the files off GitHub by hand. After that you type /skill-name in the chat.
Does it cost anything?
No. The whole library is free and open source (MIT) on GitHub. Read it, fork it, adapt it to your own work.
Do skills blow up token usage?
The opposite. Claude reads only each skill's short header until that skill is genuinely needed, so you can install the entire library without hurting the context window.
How do master skills differ from agents?
An agent (a single skill) does one job well: scoring, a reply audit, offers. A master skill is a multi-step system that orchestrates a dozen sub-skills into a finished result — a full GTM strategy built from a firm's website, for instance.
Can I adapt this to my own firm?
Yes — the code is open, tune it to your ICP and your process. And if you want me to put the system into your firm end to end — book a call.