cold-email-writer · diff
git:20260723.de898c0 to git:20260906.85504a9
34 added, 19 removed. Audit A to A.
---
name: cold-email-writer
- description: Write personalized cold emails that earn replies — research-based hooks, clear relevance, low-friction asks. Use for outbound prospecting messages.
+ description: Write personalized cold emails that earn replies - research-based hooks, clear relevance, low-friction asks. Use for outbound prospecting messages.
metadata: {"navin":{"emoji":"🧊","category":"sales"}}
---
# Cold Email Writer
- ## Overview
-
A cold email earns a reply when it proves relevance in the first line and asks for something easy. Personalization is research, not {{first_name}}.
- ## Anatomy (60–120 words max)
+ ## When to use
- 1. **Subject** — 2–5 words, lowercase-casual or specific ("question migration data", "SAFA → vous ?") — no clickbait
- 2. **Line 1** — the observation that proves you did homework (their news, job posting, public project, tech choice)
- 3. **The bridge** — that observation → the problem you solve, one sentence
- 4. **Proof** — one relevant result with a number ("aidé X à faire Y en Z semaines")
- 5. **Ask** — low friction: a question they can answer in one line, or "ça vaut 15 minutes ?" — never "book a call on my calendar" cold
+ - Outbound first-touch and sequence email copy
+ - LinkedIn note copy when asked (keep under platform limits)
+ ## When not to use
+
+ - Warm intros where a full proposal is needed
+ - Spam blasts / purchased lists
+
+ ## Anatomy (60-120 words)
+
+ 1. **Subject** - 2-5 words, specific - no clickbait
+ 2. **Line 1** - observation that proves homework (news, job post, project)
+ 3. **Bridge** - observation → problem you solve (one sentence)
+ 4. **Proof** - one relevant result with a real number you can defend
+ 5. **Ask** - low friction question or 15-minute check - not a hard calendar CTA cold
+
## Personalization tiers
| Tier | Effort | Use for |
|------|--------|---------|
- | Researched (unique first line, custom angle) | 10–15 min | A-leads, big accounts |
- | Segmented (shared pain per segment, real examples) | template + 2 min | B-leads |
- | Generic blast | never | — |
+ | Researched | 10-15 min | A-leads |
+ | Segmented | template + 2 min | B-leads |
+ | Generic blast | never | - |
## Workflow
- 1. Input: lead + research (`account-research` for A-leads).
- 2. Write 2 variants; ruthless cut to <120 words; mobile-preview check (subject + first line).
- 3. Plan the sequence with `outreach-sequencer` (this skill writes; that one schedules).
- 4. Human validation before sending (see `human-approval`); track replies to learn which hooks work.
+ 1. Input: lead + `account-research` / `buying-signals` for A-leads.
+ 2. Write 2 variants; cut to <120 words; check mobile (subject + first line).
+ 3. Plan sequence with `outreach-sequencer`.
+ 4. Human validation before send; track which hooks win.
## Rules
- - Never fake familiarity ("j'adore ce que vous faites") or fake urgency.
- - One CTA. No attachments. No "je me permets".
- - Compliance: identify yourself/company honestly; opt-out respected instantly.
+ - Never fake familiarity or fake urgency.
+ - One CTA. No attachments on first touch. No "je me permets".
+ - Do not invent case-study metrics; use real ones or omit proof.
+ - Compliance: identify yourself/company; honor opt-out.
+
+ ## Anti-patterns
+
+ - Long brochures
+ - "I help companies like yours" with no specificity
+ - Fake "saw you were looking at our site" intent lies