sequence-builder

name	sequence-builder
version	1.0.0
description	Turns an outreach brief + target segment into a ready-to-paste cold sequence for Lemlist, Smartlead, Instantly, Reply.io, Woodpecker, or Apollo sequences. Produces day-by-day steps with subject lines, bodies, merge-tag syntax correct for the chosen tool, A/B variants on the hero step, and a cost estimate at the list level. Pairs with outreach-expert (strategy) and crm-sync-expert (delivery). Proactively invoke on "write a sequence for this list", "build a cold email campaign", "give me a Lemlist sequence", "generate variants", or after build-prospect-list when the user mentions sending.
benefits-from	["richapi-gtm","outreach-expert","build-prospect-list","enrich-and-verify","list-hygiene","crm-sync-expert","scheduled-workflow"]
allowed-tools	["Bash","Read","Write","Edit","AskUserQuestion"]
triggers	["write a sequence","build a cold campaign","Lemlist sequence","Smartlead sequence","Instantly sequence","Reply.io sequence","generate variants","cold email copy"]

sequence-builder

Outputs a complete, paste-ready cold sequence for one sending tool. Not a copywriter — a playbook executor that takes the strategic decisions (segment, pain, proof, tool) and produces every field the tool needs.

Preamble

~/.claude/skills/richapi-gtm-skills/bin/richapi-skills-preflight

Phase 0 — gather the brief

Ask these in one AskUserQuestion batch — don't drip them out. Required:

Slot	Example
Sending tool	Lemlist / Smartlead / Instantly / Reply.io / Woodpecker / Apollo
Target segment	"VP Sales at 200-500 SaaS in NYC / SF"
Pain / trigger	"Spending too many manual hours on LinkedIn prospecting"
Offer	"A Claude skill pack that turns the Richapi MCP into opinionated GTM workflows"
Proof asset	Case study, data point, logo, or customer quote — one thing
CTA style	Soft ask ("worth a 10-min call?") / hard ask ("15 min Thursday?") / value-first ("here's a free tool")
Personalization depth	L0 firstName only / L1 company hook / L2 dynamic image / L3 hand-researched
Sequence length	2 steps (high-ICP only) / 4 steps (default) / 6 steps (warm lists)
Language + timezone	"English, US East"

Optional:

• Sender persona (founder vs SDR — the copy voice differs a lot)
• Brand voice samples to match
• Any hard no-go words (industry-sensitive)

If the user skipped outreach-expert first, flag it: "I can build this now, but if you haven't decided on sending tool / cadence / deliverability setup, outreach-expert is worth 5 minutes first."

Phase 1 — skeleton structure

Default 4-step cold sequence. Adjust step count based on the brief.

Step 1  Day 0   Intro + hook + one-sentence why + soft CTA
Step 2  Day 3-4 Value-add: a resource, stat, or question that proves relevance
Step 3  Day 7-8 Social proof: customer quote, case-study snippet, or logo cluster
Step 4  Day 12-14 Break-up: "Should I stop following up?"

Send-time defaults per step: Tue-Thu, 08:30-10:30 recipient-local. Tool-specific schedulers take this automatically.

Phase 2 — copy — the templates

Step 1 rules

• ≤ 100 words
• One CTA
• No image, no tracking pixel on step 1 (deliverability)
• Subject line: 3-5 words, lowercase, no brackets, no emoji
• Personalization slot: first line is always specific to them

Step 1 template

Subject: {{personalized_subject}}

Hi {{firstName | "there"}},

{{first_line_personalization — one sentence that proves this isn't templated to 10k people}}

{{value_prop — one sentence specific to their segment}}

{{soft_CTA — a question, not a demand}}

— {{senderFirstName}}

Where the slots come from:

• personalized_subject: pick from ≤5 variants, tested weekly; keep hero + 2 A/B variants below
• first_line_personalization: a recent event (job change, post, hire, fundraise) — account-research and evidence-score surface these signals
• value_prop: specific to segment, never to the individual
• soft_CTA: "Worth a 10-min look?" beats "Can we hop on a call Thursday at 2?"

Step 2 template

Subject: re: {{step_1_subject}}

Hi {{firstName | "there"}},

{{bridge — one sentence: "wanted to share something specific to {{company | "your team"}}"}}

{{value_asset — link to a one-page breakdown, a data point, or a quick framework}}

{{light_ask — "Curious if this maps to what you're seeing?"}}

— {{senderFirstName}}

Step 3 template

Subject: how {{relatable_peer_company}} used this

Hi {{firstName | "there"}},

{{proof_one_liner — "{{peer_company}} was running into {{their_pain}}. Three weeks in: {{outcome_number}}."}}

{{transition_to_them — "Thought of you because {{company | "your team"}} might have the same shape."}}

{{cta — "15 min next Tuesday? Can send a deeper breakdown first if useful."}}

— {{senderFirstName}}

Step 4 (break-up) template

Subject: closing the loop

Hi {{firstName | "there"}},

Haven't heard back — that's fine. Just didn't want to keep showing up uninvited.

If {{offer_summary}} isn't a fit right now, a quick "not now" is all I need and I'll stop.

— {{senderFirstName}}

Phase 3 — subject-line variants

Always produce hero + 2 variants on the step-1 subject. Label them for the user to A/B.

Example (SaaS Directors+ pain = manual prospecting):

• Hero: quick sanity check
• Variant A: 3 hours back for your SDRs
• Variant B: re: your stack

Rules:

• None longer than 40 chars
• No !, no [brackets], no capitalized hooks
• Don't repeat company names already in the "From" line
• re: outperforms fresh subjects — use where honest (reply continuity)

Phase 4 — tool-specific export

Emit the sequence in the syntax of the user's chosen sending tool. Pick the right one.

Lemlist

Uses Liquid-like syntax for variables + a special {{icebreaker}} field for first-line personalization.

Subject: {{subject_line}}

Hi {{firstName | default: "there"}},

{{icebreaker}}

We help {{companyType | default: "teams like yours"}} {{valueProp}}.

{{cta}}

— {{senderFirstName}}

Lemlist has an "AI icebreaker" feature; for L1/L2 personalization depth, you can populate icebreaker upstream (per-lead in the CSV) or let Lemlist's AI generate it from a prompt.

Smartlead

Uses spintax for random variation + standard merge tags.

Subject: {{subject_spin | "{quick sanity check|re: your stack|3 hours back for your SDRs}"}}

Hi {{first_name}},

{intro_variant | "Saw|Came across|Noticed"} {{company}} just {{recent_event}}.

{{value_prop}}

{cta_variant | "Worth 10 min?|Would a quick look help?|Open to a call?"}

— {{sender_first_name}}

Spintax runs at send time; each recipient gets one deterministic variant.

Instantly

Standard {{variable}} merge tags. Supports custom fields from the uploaded CSV.

Subject: {{subject}}

Hi {{firstName}},

{{customField1_firstLine}}

{{value_prop_static}}

{{cta_static}}

— {{sender}}

Instantly is cleanest when you pre-compute first-line personalization upstream and attach it as a custom field on the CSV.

Reply.io

Supports merge tags + conditional blocks. Multi-channel tool — sequence includes LinkedIn / email / call steps.

Subject: {{Subject}}

Hi {{FirstName}},

{{if CompanySize}}For teams your size, {{/if}}{{ValueProp}}.

{{CTA}}

— {{Sender.FirstName}}

Conditionals help for segmented lists.

Woodpecker

Standard merge tags. EU-hosted; mind GDPR lawful-basis block if targeting EU.

Subject: {{SUBJECT}}

Hi {{FIRST_NAME}},

{{SNIPPET1}}

{{CTA}}

— {{FROM_FIRST_NAME}}

Uses ALL_CAPS tokens — distinct from the others.

Apollo sequences

Apollo's sequence tool uses {{fieldName}} matching its own data model (firstName, companyName, titleKeyword). If the user hasn't enriched via Apollo, Apollo won't have fields your CSV filled in — push via their CSV importer first.

Phase 5 — A/B testing plan

If the user wants multivariate testing:

• A/B only step 1. Steps 2-4 carry follow-up logic; testing them is noise on small lists.
• Minimum 200 sends per variant before declaring a winner. Below that, you're reading noise.
• Test one variable at a time: subject, OR first-line, OR CTA. Never two.
• Hero metric: positive reply rate. Not opens (post-MPP they're garbage — see outreach-expert).

Phase 6 — cost math at the list level

Before exporting, give the user a line-item cost forecast. The list needs to be through build-prospect-list + enrich-and-verify + list-hygiene before sending.

List size:            500 leads
Data pipeline cost:
  build-prospect-list (lead_search, 20 pages) : 250 credits
  enrich-and-verify   (bulk enrich + emails)  : 800 credits (approx, 70% hit)
  list-hygiene        (classify + verify)     : 250 credits
Data total:                                     ~1300 credits

Sending tool cost:     per-seat subscription — not Richapi credits

Make this clear — Richapi credits are for data; the sending tool charges separately per seat. Don't conflate the two.

Phase 7 — scheduling the campaign launch

After producing the sequence, offer:

Want me to schedule this as a recurring ICP refresh? Every Monday 05:00, pull 100 fresh leads matching your ICP through the full data pipeline, and drop the CSV ready-to-import into your sending tool. → scheduled-workflow

Most cold sequences run on a new-leads-per-week cadence. The sequence copy is static; the list is what needs refreshing. Scheduling this loop is the unlock.

Phase 8 — output structure

Deliver in one message, in this order:

1. One-line summary: "4-step sequence for , for , CTA-style: ."
2. Hero + 2 variants for step-1 subject.
3. Full copy for steps 1-4 in the chosen tool's native syntax.
4. List of required custom fields / merge tags the user must have on their CSV.
5. Cost forecast at the list level.
6. Scheduling offer (always).

If the user asked for the response as a file, write to ./sequences/<segment-slug>-<tool>.md (creating ./sequences/ if absent).

Anti-patterns

• Writing copy longer than 100 words in step 1. 50-80 is the sweet spot.
• Putting the proof in step 1. Prove relevance first; proof comes in step 3.
• Using the same CTA ("Are you open to a call?") on every step. Each step shifts the ask shape.
• Offering multiple CTAs in one email. Pick one. Always.
• Adding emoji to subject lines for "pop". It's a spam tell; filter hates it.
• Hard-coding a specific sending day/time instead of "Tue-Thu, 08:30-10:30 recipient-local" — the tool schedulers handle this for you.
• Writing "I hope this email finds you well" — pattern-matched as cold-email spam by humans and filters.
• Using {{firstName}} without a default — empty merge in subject = instant trash.
• Testing six variants at 30 sends each. Not significant. Stop.

Handoff patterns

• "This list needs hygiene before it sends" → list-hygiene
• "Push this list into my CRM / sending tool" → crm-export or crm-sync-expert
• "Run this ICP refresh weekly" → scheduled-workflow
• "Revisit the strategy — am I even sending this the right way?" → outreach-expert
• "Rank the list before send — who should hear from us first?" → evidence-score

Cost

This skill doesn't spend Richapi credits. It transforms an already-enriched list into a sequence. All the data spend happened upstream.