cold-email-skills

Your agent can write code. Now it can write cold email that gets replies.

10 free, open-source Agent Skills for planning and writing cold-email campaigns, extracted from the playbook we run our agency on. MIT licensed. No account, no API keys, no telemetry. Everything is written to plain markdown files on your machine.

npx skills add Ken-Technology/cold-email-skills

Works with Claude Code, Codex, Cursor, and any other agent that supports Agent Skills.

Where this comes from

Ken started as a done-for-you cold email agency. These are not prompts written for a repo - they are the production skills our team runs for clients, ported to work anywhere. Across all client campaigns, this playbook averages a 3% reply rate against a 0.8% industry average and 7 meetings per 10k emails against the industry's 1, with copy frameworks tested across 500k+ sent emails.

We open-sourced the strategy layer because the playbook was never the moat. Execution is.

What this looks like

You say:

Plan a cold email campaign for my devtools startup. We sell a CI cost dashboard to engineering leaders at Series A-C companies.

Your agent runs the cold-email-campaign skill, asks a handful of questions about your offer, ICP, and proof points, and then works through research, targeting, qualification, segmentation, per-segment strategy, copywriting, review, and personalization prompts. Twenty minutes later you have:

cold-email/acme-outbound/
├── research.md          # what you sell, to whom, and why they care
├── plan.md              # 1 ICP + 1 offer + N segments
├── search-strategy.md   # portable list filters + a bring-your-own-list CSV shape
├── qualification.md     # AI prompts that judge prospect fit
├── segmentation.md      # rules that route each prospect to a segment
├── 0 - default/
│   ├── strategy.md      # sequence blueprint + personalization variables
│   ├── emails_v2.md     # reviewed, ready-to-send copy
│   └── prompts.md       # personalization prompts you can run in any LLM
└── 1 - series-a-eng-leaders/
    └── ...

Every email is drafted by email-copywriting, then scored by email-review against hard copy rules and rewritten until it passes. You get the reviewed version, not the first draft.

Install

Ranked by friction:

1. Have your agent do it. Paste this into Claude Code, Codex, or Cursor:

Install the skills from github.com/Ken-Technology/cold-email-skills, then plan a cold email campaign for me.

2. Install the full set:

npx skills add Ken-Technology/cold-email-skills

3. Install a single skill:

npx skills add Ken-Technology/cold-email-skills/email-copywriting

Each skill is a plain markdown file. Worst case, open the SKILL.md and paste it into any LLM.

The skills

Skill What you get
cold-email-campaign The orchestrator. Runs the whole playbook end to end and stops at ready-to-send files.
client-research A research doc synthesized from your notes, meeting transcripts, and website.
search-strategy Your ICP turned into portable search filters for Apollo, Sales Nav, Clay, or any list tool.
qualification AI prompts that evaluate whether a prospect actually fits your audience.
segmentation AI prompts that divide a prospect list into campaign segments.
campaign-strategy A per-segment sequence blueprint with AI personalization variables.
email-copywriting The actual emails: hooks, body, CTAs, signatures, personalization placeholders.
email-review A hard-rule review pass that scores copy 1-5 and iterates until it clears 4.5.
prompt-writer Personalization prompts you can run in any personalization tool or LLM.
lead-magnet Lead magnet ideas and designs for click-based and reply-based campaigns.

Run the orchestrator for the full pipeline, or any skill on its own - they all read and write plain files.

What these skills do not do

They plan and write. They do not scrape contacts, verify emails, warm up inboxes, or send anything. You bring your own prospect list (search-strategy documents the CSV shape) and your own sending tool - Instantly, Smartlead, Lemlist, a Gmail sequence, whatever you already use.

That is a feature: no account, no API keys, and nothing leaves your machine.

When you outgrow the files

The skills stop where infrastructure starts. Ken is the same playbook running with the machinery attached:

The skills give you Ken adds
Portable search filters A 280M+ contact database that runs them, with verified emails and phones
Qualification and segmentation prompts The same prompts executed across your whole list automatically
Personalization prompts to run by hand Live AI personalization on every outgoing email
Ready-to-send copy Sending infrastructure, deliverability, and reply tracking
  • Try Ken AI - bring the campaign folder these skills produced; it maps 1:1.
  • Ken Daily - not ready to send at scale? Get 10 verified leads in your inbox every morning. Free forever, no card.

FAQ

Why give the playbook away? Because strategy was never the hard part to defend. Contact data, verification, deliverability, and personalization at scale are. Open-sourcing the playbook costs us nothing and raises the floor for everyone's cold email - including your prospects' inboxes.

Where does my data go? Nowhere. The skills write local files. No server, no telemetry, no account.

Which agents does this work with? Anything that supports Agent Skills via skills.sh: Claude Code, Codex, Cursor, and others. The skills are plain markdown, so they degrade gracefully to copy-paste.

Can I change the rules? Fork it. The copy rules in email-review, the sequence blueprints in campaign-strategy, the CSV shape in search-strategy - it is all markdown. PRs welcome if your changes would help everyone.

Contributing

Issues and PRs welcome. If a skill produced weak output, open an issue with the prompt you used and the artifact it wrote - that is the most useful bug report for a skills repo.

License

MIT