agentleFS
megandmartinmegandmartin/agent-skills-repo/skills/business-ops/cold-outreach-sequencer/SKILL.md

cold-outreach-sequencer

Skill4 starsChanged 4 months agoMITView on GitHub

Turn a prospect dossier into a 3-touch outreach sequence — cold email, LinkedIn message, and follow-up — with personalization rules and anti-spam guardrails. Use when the user asks to "write outreach", "draft a cold email", "build a sequence", "follow up with this lead", or has a dossier and wants first contact. Always delivers drafts; never sends anything. Don't use for researching the prospect — run lead-research-dossier first.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

Cold Outreach Sequencer

Converts a researched dossier into three short, personalized touches a busy prospect might actually answer. Hard rule: this skill produces drafts only. It never connects to an email account, never auto-sends, never schedules sends — the human presses every send button.

When to Use

  • User has a dossier (ideally from lead-research-dossier) and wants first-contact messaging.
  • User wants a cold email, a LinkedIn touch, and a follow-up as one coherent sequence.
  • User wants an existing cold draft tightened up against the guardrails below.
  • Not for: researching the prospect — run lead-research-dossier first. Not for logging the touch — update pipeline-tracker after sending. Not for bulk spray-and-pray blasts: this skill personalizes one prospect at a time and will refuse mass-merge requests.

Quick Reference

ActionCommand / Call
Input neededDossier: hooks, contact + role, ICP fit evidence, user's offer + proof point
Touch 1Cold email — ≤120 words, 1 hook, 1 low-friction CTA
Touch 2 (+2–3 days)LinkedIn — ≤60 words, references a different hook, no link
Touch 3 (+4–5 days)Follow-up email — ≤80 words, adds new value, clean break-up line
Personalization checkHook passes "couldn't be sent to any other company" test
Spam checkSee guardrails in step 5

Procedure

  1. Precheck — confirm you have: a dossier (or at minimum 2 verified personalization hooks), the contact's name + role, the user's offer in one sentence, and one proof point (metric, customer, result). Missing hooks? Stop and route to lead-research-dossier — outreach without research is spam.
  2. Frame the angle — pick the single dossier signal that best connects prospect pain to the user's offer. One angle for the whole sequence; each touch expresses it differently.
  3. Draft Touch 1 (email) — subject ≤5 words, lowercase-natural, no clickbait. Body: hook (1 sentence) → relevance bridge (1–2 sentences) → proof point → interest-based CTA ("worth a look?" not "grab 30 minutes on my calendar"). ≤120 words.
  4. Draft Touches 2 and 3 — LinkedIn touch uses a different hook, conversational, no links. Follow-up email adds something new (insight, resource, sharper question) — never "just bumping this". End touch 3 with a polite close-the-loop line.
  5. Guardrail pass — enforce all of: no deception ("Re:" fakery, fake forwards, false familiarity), no fabricated claims or invented mutual connections, mention how you found them if natural, every email carries an opt-out line, no ALL-CAPS/spam-trigger phrasing ("act now", "guaranteed"), no more than 3 touches without a reply, and flag to the user that commercial email must comply with local law (CAN-SPAM/GDPR/PDPA — physical address + working unsubscribe on commercial sends).
  6. Deliver as drafts — output the template. State plainly: "These are drafts. Review, personalize the bracketed bits if any, and send from your own client." If the user asks you to send automatically, decline and restate the drafts-only rule.
  7. Handoff — remind the user to log the touch in pipeline-tracker (stage → Contacted, next_action → Touch 2 date).

Output Template

# Outreach Sequence: <Contact> @ <Company>
Angle: <one sentence — the signal→offer connection>

## Touch 1 — Email (Day 0)
Subject: <≤5 words>
<body ≤120 words, ends with opt-out line>

## Touch 2 — LinkedIn (Day 2–3)
<≤60 words, no links>

## Touch 3 — Follow-up Email (Day 6–8)
Subject: Re: <touch 1 subject>
<body ≤80 words, new value + break-up line>

Send checklist: [ ] correct name/company  [ ] hooks still current  [ ] sent manually by you

Pitfalls

  • Template smell — if swapping the company name leaves the email intact, personalization failed. Recovery: rewrite the opening sentence around a hook that references something only this prospect did, dated within ~90 days.
  • Stale hook — the "recent" funding round was 14 months ago; referencing it signals lazy research. Recovery: check hook dates in the dossier before drafting; if all hooks are old, refresh via lead-research-dossier first.
  • CTA too heavy — asking a stranger for 30 minutes in touch 1 kills reply rates. Recovery: downgrade to an interest question; the meeting ask belongs in the reply thread, not the cold open.
  • User asks the agent to send — tempting, out of bounds. Recovery: decline, explain the drafts-only guardrail (deliverability, consent, legal exposure), and hand over paste-ready drafts instead.

Verification

  • Exactly 3 touches, each within its word limit
  • Each touch uses a hook; touches 1 and 2 use different hooks
  • No fabricated claims, fake "Re:", or invented familiarity anywhere
  • Emails include an opt-out line; compliance note surfaced to the user
  • Delivered as drafts with an explicit "you send these manually" statement
  • User pointed to pipeline-tracker for logging the touch

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.