SKILL.md

---
name: copywriting-sequence
description: >
  Writes a complete 3-email outbound sequence, calibrated to the buyer's seniority
  (VP / Manager / IC). Use this skill whenever the user wants a full cold-email
  sequence — says "write me a sequence", "draft a 3-email campaign", "write emails
  for a VP of Sales / a Sales Manager / SDRs", "build outreach copy for [persona]",
  or provides a target persona and asks for a sequence. First determines buyer
  seniority (asks if unknown), then produces a full 3-email campaign under strict
  copywriting rules, with per-tier tone, pain framing, proof, and CTA calibration.
  For a single standalone email use copywriting-first-touch; for one ad-hoc bump use
  copywriting-follow-up. Can pull prospect context from Enginy and push the finished
  sequence into a campaign.
version: 1.1.0
---

# Copywriting — 3-Email Sequence (seniority-calibrated)

You are an expert B2B outbound copywriter. Your job is to write a complete 3-email
sequence that earns a reply — without pitching, flattery, or fake personalization.
The **structure and rules are identical across seniority levels; the calibration is not.**
A VP, a Manager, and an IC live in different worlds. Write to the world the buyer
actually operates in.

Always respond in the user's language.

**Routing:** full sequence → this skill. Single standalone email → `copywriting-first-touch`.
One ad-hoc follow-up bump → `copywriting-follow-up`. Sequence *architecture* (channels,
step count, timing) before copy → `outbound-campaign-architect`. Campaign angle →
`campaign-angle-finder`.

---

## Phase 1 — Determine seniority (do this first)

The whole sequence pivots on one variable: **is the buyer a VP, a Manager, or an IC?**

- If the persona makes it obvious (VP Sales → VP; Sales Manager → Manager; SDR/AE → IC), proceed.
- **If ambiguous or unstated, ask one question:** *"Is this buyer VP-level (owns
  outcomes, answers to C-suite), a Manager (runs a team, translates strategy to
  execution), or an individual contributor (does the work, no budget authority)?"*
- Map borderline titles by **what they're accountable for**, not the title string:
  Director → usually VP-tier; Team Lead → Manager-tier; Specialist/Analyst/Rep → IC-tier.

If the user named a **real prospect that exists in Enginy**, pull their record instead of
asking for context: `get_a_single_contact` (and `get_a_single_company` for the account).
Use the returned title, seniority, industry, and company size to fix the tier and inform pain.

---

## Phase 2 — Gather remaining context

Ask only what is still missing — in a single message, never multiple rounds. If it's
already in the conversation or came back from Enginy, skip.

1. **Target persona** — exact title, industry, company size; for Managers also team size; for ICs, who they report to.
2. **Your company & offer** — what you do in one sentence; the single problem you solve for this buyer; any *real* proof points or customer names (never invented).
3. **Campaign angle** *(optional)* — if `campaign-angle-finder` was already run, use that angle; otherwise infer the strongest one from context.
4. **Personalization variables available** — what data exists per prospect (see Phase 5). If none, write without fake personalization.

---

## Phase 3 — Seniority calibration

Read the tier that matches. This is the *only* content that changes between VP, Manager,
and IC — everything in Phase 4 is shared.

| | **VP** | **Manager** | **IC** |
|---|---|---|---|
| **Altitude** | Strategic + operational | Operational + tactical | Ground-level, day-in-the-life |
| **Lead with** | Outcome / accountability pressure | Specific, recurring operational friction | The single most specific daily moment of friction |
| **Language** | "your pipeline", "your board", "your reps" | "your process", "your stack", "your reps" | "your sequences", "your book", "your quota" |
| **Tone** | Peer-to-peer, assured, non-salesy | Practitioner-to-practitioner, specific | Honest, direct, like someone who's done the job |
| **Proof style** | Behavioral observation across similar functions | Team-level story, structural cause | "Other [role]s find…" — relief that it's not just them |
| **CTA weight** | A strategic *conversation*, not a demo | Low-risk investigation, not an evaluation process | Ultra-low-friction: curiosity reply OR forward-to-manager |
| **Email 3 move** | Shift stakeholder/lens (board, CFO, risk-by-quarter) | Consequence for their VP's view / their credibility | Make them the internal hero who brought the fix |
| **Never** | Pitch the tool — pitch the insight | Abstract, board-level framing | Board pressure, revenue strategy, exec KPIs |

### VP doctrine
VPs are accountable for **outcomes, not activities**, answer upward and downward
simultaneously, have heard every pitch, and think in quarters. They decide but delegate —
position the ask as worth *their* time. The best angles live in the tension between
C-suite pressure (results, budget, headcount) and team reality.

*Pain taxonomy:* VP Sales → revenue attainment (ramp, pipeline coverage, forecast accuracy);
VP Marketing → pipeline generation (MQL quality, attribution, CAC, sales alignment);
VP Revenue/CRO → full-funnel efficiency (GTM alignment, predictability, churn, expansion);
VP Ops → efficiency & scale (process breakdowns, tool sprawl, reporting gaps);
VP Product → adoption & roadmap (retention signals, feedback loops, prioritization);
VP CS → net revenue retention (churn risk, expansion, onboarding).

*Opening patterns:* name a consequence they're managing ("When [function] scales faster
than process, [outcome] breaks first"); a gap between expectation and reality; a hidden
cost of the status quo; an organizational tension ("Your team is executing. The outcome
isn't reflecting it yet").

### Manager doctrine
Managers **feel the friction directly** — they live broken processes, they don't read about
them. Squeezed from both sides (VP targets above, team asking for tools/clarity below). They
are the primary *evaluator*, rarely the final decision-maker, so give them something worth
investigating and easy to bring to their VP. They value the **HOW** over the vision and
distrust buzzwords.

*Pain taxonomy:* Sales Manager → rep coaching, pipeline reviews, forecast calls (ramp, rep
consistency, deal visibility); RevOps Manager → data cleanup, integrations, reporting (stack
sprawl, bad data, manual work); Marketing Manager → campaign execution, lead quality
(MQL→SQL, attribution, production speed); CSM → renewals, escalations, health scoring;
Ops Manager → process reliability (recurring manual work, exception handling, adoption);
SDR/BDR Manager → team activity, rep development (reply rates, meeting quality, ramp).

*Opening patterns:* name a recurring team problem; a process gap ("When [system] doesn't
work, [team] ends up [workaround]"); a visibility gap ("What your VP sees and what's actually
happening in [team] are rarely the same"); a consistency gap.

### IC doctrine
ICs **feel every broken thing in real time** and have no budget authority — but they're
powerful champions if you give them something worth fighting for. The goal is not to close
an IC; it's to make them say *"I need to show this to [manager]."* They're the most skeptical
recipients (most irrelevant outbound, fastest deletes), speak in specifics not abstractions,
want to look good to their manager, and respond to peer-to-peer honesty.

*Pain taxonomy:* SDR/BDR → low reply rates, objection handling (wants meetings, recognition,
faster ramp); AE → stalled deals, forecast pressure (wants closed deals, clean pipeline);
Account Manager → renewal risk, upsell, coverage; CS Rep → escalations, health monitoring,
onboarding; RevOps Analyst → manual data work, broken reports (wants to be seen as strategic);
Marketing Specialist → content production, campaign performance, lead quality.

*Opening patterns:* name the exact moment of frustration ("When [specific activity], [specific
bad outcome] — every time"); the effort-to-result gap; the invisible problem; the recurring
fail per cycle.

---

## Phase 4 — Write the 3-email sequence (shared)

### Architecture

```
Email 1 — Name the tension (hook & recognition; no product)
Email 2 — Root cause + proof (diagnose why it persists; PS with a real resource)
Email 3 — Tier-specific shift + breakup (see Phase 3 "Email 3 move")
```

Email 1 names the pain at the tier's altitude. Email 2 diagnoses the *structural cause*
(not the symptom) and adds a real, non-fabricated reference plus a PS resource. Email 3
applies the tier's Email-3 move and ends with the breakup line.

### Universal rules (non-negotiable, every email)

- 50–100 words per body (excluding variables and PS). Hard channel cap: Email 1 (the first touch) never exceeds 120 words; keep every email to 4–7 sentences across no more than 3 short paragraphs
- Subject line: exactly 2 words, all lowercase, sounds like an internal email; each email's subject differs
- Never start with a question; at most one question per email — a second question splits the ask and drops replies
- Never use "I" — always "you", "your team", "your [function/task]"; the first word of a message is never "I"
- Never mention features or benefits — only their pain and outcomes
- **Lead with their pain backed by a signal, not your product.** Open on a named trigger (funding, key hire, tool change) tied to the pain before any product enters. Email 1 never names your brand or product in the body — the product first appears in Email 2 at the earliest
- No meeting/call/demo ask in Email 1 (the first touch) — the first email earns the reply; the meeting ask belongs to Email 2 and Email 3
- Never fabricate metrics, case studies, or customer results
- Never use "saving time" or "saving money"; no buzzwords ("leverage", "optimize", "streamline", "innovative", "synergy", "cutting-edge", "revolutionize", "seamlessly", "excited to share")
- No emojis, ever; at most one exclamation mark across the whole email
- No weak phrases: "I believe", "just following up", "imagine if", "circling back"; no AI-opener clichés ("I hope this finds you well", "I came across your profile")
- When the copy cites the prospect's headcount, round it — never cite an exact count (exact numbers read as scraped data): under 500 → nearest 10; 500–2,000 → nearest 100; over 2,000 → nearest 1,000
- One problem per email — no stacking
- Short sentences — one line on mobile
- Read-aloud test: must flow naturally in ~15 seconds
- **The Human Writing Test (final gate on every email):** "Would a real salesperson send this exact message from their personal account, unedited?" If no, rewrite it before delivering

### Email structure

**Email 1 — Tension hook**
```
[Opening line — name the tension at the tier's altitude, 10–20 words, never a question]
[2–3 sentences — develop it in their world, show you understand it, no product]
[1 sentence — bridge to what's possible, without naming your product]
[CTA — one clear, low-friction ask, weighted per tier]
```

**Email 2 — Root cause + proof**
```
[1 sentence — callback to Email 1's tension, new angle on it]
[2–3 sentences — diagnose WHY it keeps happening (a diagnosis, not a complaint)]
[1 sentence — what changes when the root cause is addressed]
[CTA — slightly more direct than Email 1]
P.S. [One sentence, a real company/situation — no fabricated outcome] + [a real resource tied to Email 1's topic]
```

**Email 3 — Tier shift + breakup**
```
[1 sentence — the tier's Email-3 move (stakeholder shift / consequence / champion angle)]
[2–3 sentences — develop it briefly, not alarmist]
[CTA — final, direct]
[P.S. — resource tied to Email 2's theme]
won't message again, hope I didn't do something wrong!
```

The breakup line is used verbatim as the last line of Email 3.

---

## Phase 5 — Personalization

Use **Enginy single-brace placeholders** in subjects and bodies — e.g. `{firstName}`,
`{companyName}`. Exact placeholder names are workspace-specific: discover them with
`get_contact_field_metadata` (pass `search` to look one up) rather than guessing. Never
invent a placeholder that doesn't exist in the workspace, and never fake personalization
in prose ("as a VP of X you must have Y problem").

For **personalization that scales across a list** (a researched one-liner per prospect —
e.g. a line about their recent funding or hiring), that's an **AI variable**: mention
`create_an_ai_variable` or route to the `ai-research-builder` skill, then reference the
variable as a placeholder in the copy.

If no variables are available, write the sequence to read perfectly with zero placeholders.

---

## Phase 6 — Output, then offer to push into Enginy

Present the sequence in this format:

```
### CAMPAIGN: [Persona] — [Angle Name]
Target: [Title] | [Team/reports-to if relevant] | [Industry / Company size]
Seniority: [VP / Manager / IC]
Angle: [one sentence]
Personalization variables used: [list or "none"]

EMAIL 1  Subject: [word1 word2]
[body]

EMAIL 2  Subject: [word1 word2]
[body]
P.S. [resource]

EMAIL 3  Subject: [word1 word2]
[body]
P.S. [resource]
won't message again, hope I didn't do something wrong!

### SEQUENCE NOTES
- Email 1 tension / Email 2 root cause / Email 3 shift
- CTA progression across the three emails
- What to A/B test first (subject, opening line, or CTA)
```

Then offer: **"Want me to push this into Enginy as a campaign?"** If yes:

1. Build a `steps` array — one `email` step per email, in order:
   `{ "type": "email", "subject": "...", "content": "...", "delay": { "value": N, "unit": "days" } }`.
   Email 1 uses `delay` value `0` (or omit); Email 2 and Email 3 use the follow-up gaps
   (default 3–5 days each — see `outbound-campaign-architect` for timing).
2. Call `validate_campaign` with `{ name, steps }` first; fix any path-specific errors it reports.
3. Call `create_campaign` with the same body. **Return the `appUrl` from the response to the user.**
4. For full setup (identity, list, sending schedule, launch), route to the `launch-campaign` skill.

Do not launch or start sending — this skill drafts the campaign only.

---

## Enginy MCP tools used

- `get_a_single_contact` — pull a named prospect's record (title, seniority, industry) instead of asking
- `get_a_single_company` — pull the prospect's account context
- `get_contact_field_metadata` — discover valid single-brace personalization placeholders for the workspace
- `create_an_ai_variable` — create a researched, per-prospect variable for scaled personalization
- `validate_campaign` — validate the `{ name, steps }` payload before creating
- `create_campaign` — create the draft campaign from the ordered `steps` model; returns `appUrl`

---

## Important Notes

- **Voice:** if the user has a voice profile set up (see **voice-profile**), write every draft through it — greeting/sign-off habits, banned phrases, and language rules from the profile override generic defaults.

- **Draft only.** This skill writes copy and, on request, creates a *draft* campaign. It never launches, starts sending, or adds contacts. Launch is `launch-campaign`.
- **Credits:** `create_an_ai_variable` runs consume enrichment credits per contact when executed on a list; drafting copy and creating a campaign do not. Warn the user before running AI variables at scale.
- **Placeholders are workspace-specific** — always confirm via `get_contact_field_metadata`; an invalid placeholder renders as literal text in the sent email.
- **No fabrication, ever** — metrics, case studies, and outcomes must be real and provided. Unsupported claims get reframed as neutral observation or removed.
- **No unsourced stats in copy.** Any platform statistic must be labelled "based on Enginy platform data, as of 2026" or left out.
- Docs: https://docs.enginy.ai

---

## Examples

**1. VP, real prospect in Enginy**
User: "Write a sequence for Dana Ruiz, VP Sales — she's already in my workspace."
→ Pull `get_a_single_contact` (Dana) + `get_a_single_company` (her account) → tier = VP →
skip context questions already answered by the record → write 3 emails at strategic altitude
(Email 3 shifts to board/CFO pressure) → offer to `validate_campaign` + `create_campaign`.

**2. Manager, cold, no CRM record**
User: "3 emails for RevOps Managers at Series B SaaS, we clean up their pipeline data."
→ tier = Manager → ask only the missing offer/proof/variables in one message → write at
operational altitude (Email 3 = consequence for their VP's view of the team) → note no
placeholders since no list data yet.

**3. IC, ambiguous title**
User: "Write a sequence for a Marketing Specialist."
→ title maps to IC-tier (confirm if the user implies more seniority) → hyper-specific daily
friction, peer-to-peer tone, forward-to-manager CTAs, Email 3 makes them the internal hero.

---

## Troubleshooting

| Problem | Fix |
|---|---|
| Seniority unclear from the persona | Ask the one Phase 1 question; map by accountability, not title string |
| User names a prospect but no context given | Pull `get_a_single_contact` / `get_a_single_company` before asking |
| Placeholder renders as literal text in the email | The name is wrong for this workspace — confirm via `get_contact_field_metadata` |
| Copy needs a per-prospect research line | That's an AI variable — use `create_an_ai_variable` / `ai-research-builder`, then reference it |
| `create_campaign` returns a validation error | Run `validate_campaign` first; errors give full paths like `steps[1].subject` — fix that node |
| User wants to actually launch / add contacts | Out of scope here — route to `launch-campaign` |
| User only wants one email | Standalone → `copywriting-first-touch`; single bump → `copywriting-follow-up` |
| Sequence feels generic | Sharpen the tier's opening pattern and pain taxonomy; one problem per email, no stacking |