Cold Email Maestro

Cold Email Maestro drafts personalized, deliverability-aware cold emails and the follow-up sequence to go with them. Give it a prospect (name, company, role, or a LinkedIn/website URL) and your offer, and it returns three distinct variants, a spam-risk score for each, and a proposed cadence. It's built for founders, SDRs, and solo operators who send low-volume, high-intent outreach and care more about reply rates than blast volume.

When to use it

• You're reaching out to a specific person and want the opener tied to something real about them, not "I came across your profile."
• You want to A/B test subject lines and angles before committing a sequence to your sending tool.
• You're worried a draft reads as spammy and want it checked before it hits a real inbox.
• You need a multi-touch follow-up plan (not just one email) with sensible spacing.

What it does

Given a prospect and offer, the skill runs four stages:

1. Context gathering — pulls public signals (company site, role, recent posts if a URL is supplied) to ground the personalization line. No URL? It works from whatever fields you provide.
2. Three variants — generates three angles (e.g. problem-led, social-proof-led, direct-ask) so you can test framing, each with subject + body.
3. Spam scoring — flags trigger words, link/image ratio, all-caps, excessive punctuation, and length, returning a 0–100 risk score and the specific lines dragging it down.
4. Cadence — proposes a follow-up schedule (typically 4–5 touches over ~2–3 weeks) with the intent of each touch and draft copy.

Output is plain text/Markdown you can paste into Gmail, Apollo, Instantly, or Lemlist. The skill writes copy and analysis; it does not send mail or connect to your ESP.

Usage

curl -fsSL --create-dirs -o ~/.claude/skills/cold-email-maestro/SKILL.md https://claudeskil.com/skills/cold-email-maestro/SKILL.md

# then, in chat:
"Cold email to Dana Reyes, VP Eng at Northwind (northwind.io).
We cut CI build times ~40%. Goal: book a 15-min call."

You'll get three labeled variants, a spam score per variant with flagged phrases, and a follow-up cadence. Ask "tighten variant B to under 90 words" or "make the cadence 3 touches" to iterate.

Tips & gotchas

• Feed it a URL. Personalization quality jumps sharply when it has a real page to reference versus inferring from a title.
• The spam score is heuristic (content-side signals), not a live inbox-placement test. It can't see your domain reputation, SPF/DKIM/DMARC, or sending IP — fix those separately.
• Shorter wins: variants over ~120 words tend to score worse and reply worse. Push it to trim.
• Always sanity-check the personalization claim before sending; a wrong "I saw you raised a Series B" is worse than a generic line.

FAQ

Does it send emails or schedule the cadence automatically? No. It produces copy and a recommended schedule. You paste into your own sending tool.

Can it batch-personalize a CSV of prospects? It's tuned for one prospect at a time for quality. For a list, loop it per row and review each — it won't mail-merge a spreadsheet on its own.

Will a low spam score guarantee delivery? No. It only addresses content. Deliverability also depends on domain warmup and authentication records the skill can't inspect.