Outbound Engine
The outbound block of a GTM system, assembled as a pipeline: from the signal catalogue and scoring through to prospect lists, sequences, replies and reporting. Inside is a working method with a cost per row and gates between stages.
For the owner or head of sales at an agency or service firm · free · MIT licence
What it does
Outbound breaks somewhere other than where people fix it. Copy gets rewritten when the mail is not landing. More data gets bought when the problem is timing. This pack splits the work into stages, each with its own gate, and stops you skipping the one the next stage depends on.
The claim everything rests on: find the companies that have already described their problem in public, most often in the body of their own job ad, instead of the ones that merely resemble your clients. A company that wrote something awkward about itself paid to publish it and will not take the words back.
The second decision: a signal and a data point are different things. A signal is an event with a date, it exists for only a small share of a list, and it tells you WHEN to write. A data point is a standing fact, it exists for almost everyone, and it tells you WHAT to say. A campaign built without that split quietly turns into a mass send with a first name in the opening line.
Not sure where to start
28 skills, by stage
What you get
- An order of operations with gates across all 28 skills
- A catalogue of 60 signals and 33 data points
- A cost per row on every collection pass
- A canonical prospect-list format
- The proof rule: no URL and no date means no signal
- Two scoring models instead of one blurred one
- Ready answers for 12 objection types and 2 types of hot reply
- Known traps and failure modes, written into the skills themselves
What you need
- Claude Code to run the skills (required)
- A web-scraping service with your own API key: job postings, company profiles, site crawls
- A people-search service that works from a company domain, for contact enrichment
- Email finder and validation services, for the address cascade
- Optional: Notion, your outreach platform and your CRM over MCP, so the skills read live data and write reports as pages
Without connections the skills work from pasted data: CSV, sheets, text. Paid services are needed only at the sourcing and enrichment stages, and the cost per row is spelled out inside the skills.
The rest of the system
Outbound is one block of a GTM system. Alongside it: GTM Strategy (a cold audit from a URL into a full strategy) and Content Engine (the content system for LinkedIn).
Want this system in your own firm?
Skills are tools. Putting them to work inside your firm is the job. Book 60 minutes →
How to install
Node 18+ required. The command installs the whole set of skills into the current project folder:
npx skills add victor-shulga/outbound-engine-skillsTo keep the skill at hand in every project, add -g:
npx skills add -g victor-shulga/outbound-engine-skillsIf you work in Claude Code, the pack also installs as a plugin, with two commands in the chat:
/plugin marketplace add victor-shulga/outbound-engine-skills/plugin install outbound-engine@outbound-engine-skillsThe archive holds the folder outbound-engine-skills/skills/ : move its contents into ~/.claude/skills/, and each subfolder becomes a separate skill. Restart Claude Code. Claude Desktop and claude.ai accept one skill at a time. To upload to claude.ai: unzip the ZIP — you will see 28 skill folders inside. Zip each one individually (Mac: right-click → Compress; Windows: right-click → Send to → Compressed folder), then upload each zip in claude.ai via Settings → Skills.
Download the ZIP · 220 KBOpen the GitHub repository and copy the folder skills/ to your machine. The main file is SKILL.md, and the rest is pulled in from it.
Check it works: type /signal-outbound in the chat and the skill should answer. If it does not: skills are read at session start, so restart Claude.