campaign-copy-workflow

name	campaign-copy-workflow
description	Generate the complete Phase 2 of any web reveal campaign — static email bodies + a body-aware Clay research prompt for personalization variables. Ensures bodies and research prompt stay in sync. Use when the user types /campaigncopy <partner> or asks to build campaign copy for a partner.

Campaign Copy Workflow

Generate a complete outbound email campaign package: static email bodies (for Instantly or HeyReach) plus a Clay-compatible personalization research prompt that embeds those bodies and produces merge variables that read naturally in context.

Trigger

• /campaigncopy <partner> (e.g., /campaigncopy acme-consulting)
• "campaign copy workflow acme-consulting"
• "build campaign copy for [partner]"

Prerequisites

The partner must already have Phase 1 research in place:

• clients/partner/<partner>/research/icp.md — target accounts, verticals, qualification criteria
• clients/partner/<partner>/research/personas.md — buyer tiers and titles
• clients/partner/<partner>/prompts/signal-retrieval.md — pain signals and enrichment signals

If any are missing, tell the user to run /webreveal <partner> or /partneronboard <partner> first.

Workflow

Step 1: Load Partner Context

Read these files to understand the partner's positioning, ICP, and pain signals:

clients/partner/<partner>/SKILL.md
clients/partner/<partner>/research/icp.md
clients/partner/<partner>/research/personas.md
clients/partner/<partner>/prompts/signal-retrieval.md

Step 2: Confirm Campaign Parameters

Ask the user:

1. Campaign type: 2-message email (default), 3-email sequence, LinkedIn DM sequence, or custom
2. Sending platform: Instantly (default), HeyReach, or other
3. Target persona tier: Which tier(s) from personas.md to optimize for (default: primary outreach tiers)
4. Sender name: Who signs the emails (default: Shawn)
5. Any specific angles or constraints: e.g., "focus on cloud migration only" or "ServiceNow displacement angle"

Step 3: Write Static Email Bodies

Create/overwrite clients/partner/<partner>/prompts/campaign-copy.md with:

1. Header with status, merge fields, variable source, and campaign metadata
2. Each email in a code block with subject line and body, containing merge field placeholders
3. Copy notes section documenting voice rules and variable sourcing

Voice rules (from .cursor/skills/email-copy/SKILL.md):

• Contractions, short sentences, fragments okay
• No spam triggers (leverage, synergize, cutting-edge, revolutionary, game-changer)
• No quotation marks around brand/service names
• Human tone — "Happy to send over" not "I would be pleased to furnish"
• No "Hope this finds you well" or generic openers

Merge field convention:

• {first_name}, {company_name} — from Clay enrichment data
• {icebreaker}, {pain_point} — generated by Clay research prompt
• Additional variables only if the campaign type requires them

Email structure (2-message default):

• Email 1 (Day 0): Opens with {icebreaker} — specific, personal. Then poke the bear — calls out a known-but-unspoken problem. Value offer up front (teardown, breakdown, or insight). CTA is a quick call, not passive.
• Email 2 (Day 3): Brief callback. {pain_point} as standalone paragraph — company-specific observation. Light partner positioning (proof points). CTA is a quick call with low-risk framing.

Step 4: Generate Body-Aware Research Prompt

Create/overwrite clients/partner/<partner>/prompts/personalization-research.md with a Clay-compatible prompt that:

1. Embeds the full email body text from Step 3 — so the AI understands exactly where each variable appears and crafts them to read naturally in context
2. Includes partner service categories and pain signals from the research files
3. Defines strict rules for each variable with good/bad examples
4. Outputs JSON array per prompt-formatting standard
5. Places inputs at the bottom per .cursor/rules/prompt-formatting.md

Data hierarchy — the prompt must instruct the AI to prioritize data sources in this order:

1. LinkedIn profile (about, headline, experience, skills) — best for icebreaker personalization
2. Company domain / tech stack — best for pain point signals
3. Title + industry + size inference — fallback only, flag confidence as Low

The agent searches the LinkedIn profile for personalization hooks. If the profile is thin, it falls back to domain-level signals. If domain gives nothing useful, it uses role-based inference and flags it for manual review.

Prompt size discipline — keep the prompt tight. Every section should earn its place:

• Service categories: one line each, no descriptions unless they add targeting value
• Pain signals: bullet list, no explanations — the AI knows what they mean in context
• Email bodies: include in full (non-negotiable — this is what makes variables fit naturally)
• Variable rules: brief, with 2-3 good examples. Skip bad examples if the good ones are clear enough
• No redundant instructions. If the data hierarchy section already says "don't fabricate," don't repeat it in the variable rules

The research prompt eats Clay's context window per contact. Shorter prompt = better output quality at scale.

Critical: The research prompt and email bodies must always be in sync. If you change a body, regenerate the research prompt so the embedded context matches.

Step 5: Update Partner Skill Tree

After generating both files:

1. Update clients/partner/<partner>/SKILL.md:

   • Add personalization-research.md to the Prompts table if not present
   • Update campaign-copy.md description to reflect the actual campaign type

2. Update clients/partner/<partner>/workflows/campaign-patterns.md:

   • Register the new play in the Active Plays section with target persona, trigger, and status

Step 6: Summary

Print a summary for the user:

• Files created/updated
• Merge fields used
• Next steps (paste bodies into Instantly, paste research prompt into Clay, connect Clay → Instantly sync)

File Outputs

File	Contents
clients/partner/<partner>/prompts/campaign-copy.md	Static email bodies with merge field placeholders — paste into sending platform
clients/partner/<partner>/prompts/personalization-research.md	Clay prompt — generates personalization variables that fit naturally into the bodies

Sync Rule

The two output files are a coupled pair. If either changes, the other must be regenerated. The research prompt embeds the email bodies as context — stale bodies in the prompt produce variables that don't flow in the actual emails.

When the user asks to update campaign copy, always regenerate both files.