Systems Lab

Agent skill

outbound-sequences

Designs B2B outbound sequences (cadence, channel mix, the message for each touch) and runs a deliverability pre-flight so cold email lands in the inbox without damaging the sending domain.

activeSelf-containedActs undeclared1,528 words

Filed under Outbound email.

From kristian-presso/claude-gtm-skills · 6 skill entries · 0 · pushed 2026-10-16

What it does when it runs

Designs B2B outbound sequences (cadence, channel mix, the message for each touch) and runs a deliverability pre-flight so cold email lands in the inbox without damaging the sending domain. Use when someone asks to "write a cold email sequence", "plan an outbound cadence", "how many follow-ups should I send", "why are my emails going to spam", "set up a domain for cold outreach", "check my SPF DKIM DMARC", or wants to review an existing sequence that is getting no replies.

Automated analysis of the skill and the 5 files bundled beside it. A skill’s own description is written to be selected by an agent, so it describes the job and not the dependencies.

Keys and connectors you must supply
None found.
Hosts it reaches
No third-party host appears in the skill or its bundled files.
Tool permissions it declares
No allowed-tools in the frontmatter. It does act, so it runs under whatever permissions your session already grants.
Actions present in the files
shell

Ask about outbound-sequences

Opens your assistant with this page's recorded context. Check the linked sources before relying on it.

Is this safe to install?ClaudeChatGPT
Adapt it to my stackClaudeChatGPT
What else do I need for it to workClaudeChatGPT
Rather ask a human? Talk to Cheetah
git clone --depth 1 --filter=blob:none --sparse https://github.com/kristian-presso/claude-gtm-skills.git /tmp/claude-gtm-skills
git -C /tmp/claude-gtm-skills sparse-checkout set "skills/outbound-sequences"
mkdir -p ~/.claude/skills/outbound-sequences
cp -R "/tmp/claude-gtm-skills/skills/outbound-sequences/." ~/.claude/skills/outbound-sequences/

Picked up without a restart. A project skill of the same name is shadowed by your personal one. For one repository only, swap ~/.claude/skills for .claude/skills. Claude Code docs ↗

Or take the whole library

This repo ships a .claude-plugin manifest, so Claude Code can install all 6 skills at once. Plugin skills are invoked as /<plugin>:<skill>, so they never collide with your own.

/plugin marketplace add kristian-presso/claude-gtm-skills
/plugin

The folder is the same in every client that implements the format — 46 of them — so if yours is not above, only the destination changes.

Reproduced in full from kristian-presso/claude-gtm-skills/blob/575cd8b4dc1b7da3f310dc5a55ceb1c7aaf86a67/skills/outbound-sequences/SKILL.md, which is licensed MIT (repository). 1,528 words, 14 headings.

Outbound sequences

A sequence is a planned series of touches across channels, aimed at starting one conversation with one fit prospect. This skill covers starting that conversation. What happens after a reply is a sales process, not a sequence.

Three things decide whether a sequence works, in this order:

  1. Relevance. The right person, at the right moment, about something real at their company. This is the biggest lever by far.
  2. Delivery. Copy that lands in spam does nothing, and a burned domain takes weeks to recover.
  3. Shape. Enough touches to clear the point where most replies arrive, then stop.

If the user has no list yet, or the list is a generic export with no reason to contact anyone now, start with the signal-prospecting skill and come back here.

When to use

  • Designing a new outbound sequence from scratch.
  • Reviewing a sequence that gets opens but no replies, or no opens at all.
  • Setting up a new sending domain or inbox for cold outreach.
  • Checking a programme against the law in the recipient's country before it starts.

Reviewing a sequence that is not working? Say first that open rates are not a signal of interest: mail privacy features load tracking pixels automatically, so opens are inflated and unreliable. Judge it on replies. Then check it against the touch and first-email rules below and the deliverability pre-flight.

How to design a sequence

Work through these steps with the user. Ask only for what you do not already have.

1. Pin down the target and the reason

Get three things in one or two sentences each:

  • Who. Role and company type. Narrow beats broad: one or two contacts per account.
  • Why now. The trigger for each account (a funding round, a hire, a launch). If there is no trigger, say so plainly: expect generic reply rates, and consider finding signals first.
  • The one ask. What a "yes" looks like. Usually a short reply showing interest, not a booked 30-minute demo.

2. Choose the shape

Defaults to start from, then adjust to the audience:

  • 8 to 12 touches across all channels, or 4 to 7 if email only.
  • Roughly 3 to 4 weeks in total.
  • Front-load week one (4 to 5 touches), then widen the gaps.
  • Stop at the end. Do not run past about 12 touches. Park the contact and recycle them later with a new reason.

Senior buyers and enterprise accounts usually need fewer, more considered touches over a longer window. Small-company founders reply fast or not at all.

Read references/cadence-and-channels.md when the user wants the reasoning behind these numbers, a channel split, or a day-by-day template.

3. Mix the channels

Email is the base. Add a phone call and a professional-network touch where the user can actually do them. Video is an occasional differentiator, not a channel to build on.

A sensible opening: email and a connection request on day 1, a call and voicemail on day 2 or 3, then an email that references the voicemail. Alternate channels after that.

Do not plan a channel the user will not run. A sequence with three skipped calls is worse than an honest email-only one.

4. Write each touch

Every touch must earn its place by adding something new: a different angle, a piece of proof, a useful resource, or a simpler ask. "Just bumping this" adds nothing.

Rules for the first email:

  • 50 to 125 words. Shorter is usually better.
  • Subject line under 50 characters, plain, lower case, reads like a note from a colleague.
  • Open with them, not you. One specific, accurate observation they could not receive in anyone else's inbox. A merge field with their company name is not personalisation.
  • One ask, low friction. "Worth a look?" beats "Book 30 minutes on my calendar".
  • No feature lists. One line of proof, if you have it, is enough.

A shape that works: observation, the problem it usually causes, one line of proof, a soft question.

Read references/copywriting.md when drafting or critiquing the actual words, or when the user asks for subject lines or follow-up angles.

5. Run the deliverability pre-flight

Do this before the first send on any new domain or inbox, and again whenever replies suddenly drop. Go through the checklist below and mark each item pass, fail or unknown. Do not let the user start sending with any "fail" in the first section.

6. Check the law for where the recipients are

The recipient's country decides which rules apply, not the sender's. Opt-out regimes (the US) and opt-in regimes (Canada) differ sharply, and the EU and UK sit in between. Build for the strictest regime the list touches.

Read references/legal.md before the first send into any new country. It is a reference summary, not legal advice, and it says so.

7. Agree how it will be judged

Before launch, agree the numbers that matter and the size of test that would change the plan. Reply rate, positive replies and meetings booked, tracked per step. Not opens.

Read references/metrics-and-testing.md for benchmark ranges and how to run an A/B test that means something.

Deliverability pre-flight checklist

As of 16/10/2026. Mailbox-provider rules tighten over time: verify at Google's email sender guidelines (support.google.com/a/answer/81126), Yahoo's sender best practices (senders.yahooinc.com/best-practices) and Microsoft's Outlook.com high-volume sender requirements (Microsoft Defender for Office 365 blog, updated 30/04/2025) before relying on any threshold. Google, Yahoo and Microsoft all apply the bulk rules at about 5,000 messages a day to their users, and Microsoft rejects non-compliant mail outright.

Must pass before sending (firm requirements):

  • Separate sending domain. Cold email goes from a look-alike domain, never the main business domain, so a reputation hit cannot reach invoices and normal mail.
  • SPF published on the sending domain.
  • DKIM signing on, with a 2048-bit key where the provider allows it (1024-bit is the minimum).
  • DMARC record present, at least p=none, with the From domain aligned to the SPF or DKIM domain. Plan to move towards quarantine once reports are clean.
  • SPF and DKIM both pass, not just one, if any recipients are on Outlook.com, Hotmail or Live: Microsoft requires both for bulk senders and rejects failures with 550 5.7.515.
  • Working unsubscribe in every message, honoured within two days. For bulk mail, add one-click unsubscribe headers (List-Unsubscribe and List-Unsubscribe-Post: List-Unsubscribe=One-Click).
  • Honest sender identity. Real name, a reply-to that is read, and a postal address where the recipient's law requires one.

Should pass (practitioner consensus, softer numbers):

  • Warm-up done. New inboxes ramp from about 5 to 10 sends a day over 4 to 6 weeks before full cold volume (rule of thumb, not a provider rule).
  • Per-inbox volume capped at roughly 10 to 20 cold sends a day, counting warm-up traffic. Scale with more inboxes (3 to 5 per domain) and more domains, never by pushing one inbox harder.
  • List verified. Addresses checked before sending; expected bounce rate low.
  • Complaint rate watched in the provider's postmaster tools. Keep it under 0.1%; at 0.3% filtering starts (Google and Yahoo; Microsoft publishes no threshold).
  • Bounces and non-engagers removed after each batch.

Read references/deliverability.md when the user is setting up DNS, diagnosing a spam problem, or planning volume above a few hundred sends a day.

Guardrails

  • Relevance over volume. When the user wants to widen the list, push to tighten it first.
  • Benchmarks in this skill are vendor-reported ranges, not facts. Say so when quoting them, and tell the user to re-baseline against their own numbers.
  • Legal content is reference, not advice. Point to qualified advice before a real programme in a new jurisdiction.
  • Never draft copy that misrepresents who is sending, invents a prior relationship, or uses a fake "Re:" subject line.

Example

A founder selling scheduling software to physiotherapy clinics wants a sequence for 40 clinics that have posted a job for a receptionist in the last fortnight.

  • Shape: 7 touches over 18 days. Email day 1, call day 3, email day 4, connection request day 6, email day 9, call day 12, short break-up email day 18.
  • First email (74 words): subject "front desk hire". Opens with the job post ("Saw you're hiring a receptionist at the Leeds clinic"), names the usual cause (phones and rebooking eating the front desk's day), one line of proof from a similar clinic, then "Worth seeing before you fill the role?"
  • Pre-flight: new domain registered three weeks ago, so warm-up is not finished. Recommendation: start at 10 a day from two inboxes, reach the full 40 over the first week, and send the calls from day one because they do not touch the domain.

Configuration

  • Sending set-up: the user's sending domain(s), inbox count and sending tool, so the pre-flight can be checked against what actually exists.
  • Recipient countries: decides which legal regime the sequence is built for.

Files bundled with it

These load only when the skill asks for them, so they cost nothing until it runs.

Other skills for the same job

Different authors, same problem. Matched on the words in the skill name, across every library in the catalogue except this one.

Need help setting it up?

This page tells you what outbound-sequences does and what it needs. Cheetah builds the agent setup it runs inside: data, CRM, sequencing and the guardrails.

Book a call →

The directory stays free. There is nothing gated behind this.

Build my GTM engine

Tell us what you want to improve. Fedor reviews the workflow and replies by email when there is a useful fit.

Fedor reviews every request personally. If there is a useful fit, he will reply by email.