Cold Email Body

Cold email body

Audit and draft cold email bodies only. Read-only analysis and draft text — never send, schedule, or enable tracking. Prefer plain-text, single soft CTA, signal-led personalization.

Default to the quick shape when the user asks for copy, a short draft, or "just the email." Use the full audit shape only for diagnosis, scoring, or when the user asks why something underperforms. Render only the selected shape. A single short caveat line is allowed when proof or a critical input is missing.

For a full-email request, use Complete mode: establish the offer/CTA first, draft the body, then finalize the subject. Apply adjacent offer/subject skills when available, but assemble one user-facing answer from their results rather than emitting three separate audit reports. Never block if they are not installed.

Core rules (encode verbatim)

Length and readability

• 50–100 words / 3–4 sentences, 3rd–5th grade reading level, one phone screen.

Personalization

• Signal-led personalization + removal test (delete personalized line; if email still stands it was decorative).
• Signal-led ≈ 11% reply vs generic ≈ 2% (~5×).

All performance figures in this skill are directional benchmarks, not promises or universal causal effects. Use them as priors; the user's own campaign data, ICP, list quality, offer, and sender reputation take precedence.

Structure

• Default: Signal → Problem+Solution → Proof → soft CTA
• Alternates: PAS / BAB — see [references/frameworks.md](references/frameworks.md)
• You/your, not we/our/I-centered product tours
• Exactly one soft interest CTA (single-ask ~9.1% vs multi ~2.4%)

Deliverability hygiene

• Plain text, minimal signature
• Open-tracking pixel OFF and link tracking off on cold sends (+2–10 pts placement)
• Zero links/images/attachments on touch 1
• Poke-the-bear question option
• Spintax = deliverability insurance, not reply booster

Banned

• “I hope this finds you well” (−47% replies)
• synergy / leverage / best-in-class
• circle back / just checking in / just bumping this
• quick call
• calendar link on touch 1
• congrats-opener flattery

Lists: [references/banned-phrases.md](references/banned-phrases.md). Scoring: [references/body-scoring-rubric.md](references/body-scoring-rubric.md). Examples: [references/examples.md](references/examples.md).

Workflow

1. Collect: ICP, signal, problem, offer, proof, draft (if any).
2. Mode: Audit | Draft | Complete.
3. For full responses, score /100 with the body rubric and run the

personalization removal test.

1. Enforce one soft CTA; strip links/tracking advice into Findings.
2. In Complete mode, draft 2–3 subject options after the body; keep the body

as the primary artifact and emit one assembled answer.

1. Emit output contract; draft rewrite when Mode is Draft or Audit finds

· 
````

Quick Complete shape:

````markdown
### Subject options
1. ``
2. ``
3. ``

### Email
```text

·

````

Output contract

GitHub-flavored markdown only. Blank line between blocks. Marks: ✓ · ✗ · ·. Lead ### Score with a fenced text meter: Score /100 · ; bar = █×round(total/100×20) then ░ to width 20 (monospace glance; no colored emoji).

Shape

````markdown

Cold email body

Mode

Score

Score  ███████████████░░░░░  78/100 · solid

• Total:
• Length/readability:
• Personalization (removal test):
• Structure:
• CTA:
• Hygiene (links/tracking/plain text):
• Language (banned/you-voice):
• Band:

Findings

• …

Recommended body

CTA

• ``

Subject options

1. ``
2. ``

Tracking note

• Open pixel: OFF · Link tracking: OFF · Touch-1 links/images/attachments: none

````

Safety

• Never send or enable tracking.
• Never invent personalization signals; if missing, mark · and ask.
• Never add calendar links or multi-asks on touch 1.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: outboundsync
• Source: outboundsync/skills
• License: MIT

Install and usage instructions live in the source repository linked above.