Email Campaign Brief

A Claude Agent Skill that walks a marketer through producing a tight, opinionated email campaign brief in a guided interview — then hands back a clean Markdown, Word, or HTML document.

🌐 getdyspatch.github.io/email-campaign-brief — what it does, and a one-click download.

The brief is deliberately short (1.5 pages or less). A long brief is usually a tell that the strategy isn't clear yet, so the skill pushes back on vague goals, fuzzy audiences, and multi-idea messaging until each section actually lands.

📄 Preview the output — shared/assets/brief_template.md shows the full structure of a generated brief, section by section.

🟣 Build email in Dyspatch? There's a Dyspatch-specific variant that grounds the sample layout in your account's real modules. You don't need a Dyspatch account to use this skill — the standard version is tool-neutral and works with any ESP; this variant just produces sharper briefs for teams who build in Dyspatch.

---

What it does

Ask Claude to brief an email campaign and the skill runs a short interview, section by section, producing a brief that aligns copy, design, dev, QA, legal, and stakeholders on what the work must do and must include. It covers the strategy a generic creative brief would (Background, Goal, Audience, Challenge, Opportunity, Single Minded Idea, Proof Points, Mandatories, Timing, Measurement) plus the email-specific details that cause most last-minute rework:

• Sender name & reply-to, message type, and KPIs (Primary / Secondary / Diagnostic)
• Production path, assets (existing vs. net-new), and testing/QA flags
• UTM tracking — always included, with sensible defaults proposed if you shrug
• Suppression / exclusion lists — the "wait, we sent it to who?" safeguard
• Legal & compliance by recipient region (CAN-SPAM, CASL, GDPR, CCPA, …)
• Localization — locales, translation rounds, per-locale legal and goal rollup
• Per-email mini-sections for multi-email sequences, including exit conditions

It works for a single send or a multi-email sequence (welcome series, nurture, launch + reminders, abandonment, re-engagement, and more).

What you get

• A finished brief in Markdown (default), Word (.docx), or HTML — your choice, remembered for next time.
• Optional extras you can accept or skip:
  • Recommendations pass — 1–3 research-backed, brief-specific suggestions for how the campaign could land harder.
  • Sample email — a content-block table (header, hero, body, CTAs, footer) with directional draft copy tuned to your audience.

Two variants

This repo ships two skills built from the same brief methodology:

• email-campaign-brief — the generic, tool-neutral skill. It never assumes which email-design or sending platform you use.
• email-campaign-brief-dyspatch — a variant for teams who design and produce email in Dyspatch.io. It assumes Dyspatch is your production path and, when the Dyspatch MCP server or API is connected, pulls your account's real modules/blocks so the sample layout references blocks you can actually drop into Dyspatch.

Everything else — the interview, the brief structure, compliance/localization/UTM handling, recommendations — is identical. Pick the generic skill unless you build in Dyspatch.

Installation

Each skill is a standard Agent Skill — a directory containing SKILL.md. Install whichever you want wherever your agent looks for skills.

Claude Code (skills live in ~/.claude/skills/) — use a packaged release. Download the .skill file (a zip) for the variant you want, then unzip it into your skills directory:

# Generic
unzip email-campaign-brief-<version>.skill -d ~/.claude/skills/
# Dyspatch variant
unzip email-campaign-brief-dyspatch-<version>.skill -d ~/.claude/skills/

The unpacked folder is named after the skill (email-campaign-brief or email-campaign-brief-dyspatch); leave it as-is. Restart your agent if it caches the skill list.

To work from source instead, clone the repo and copy a skill's built folder (run npm run build first), or copy skills/<name>/SKILL.md together with the shared/references and shared/assets files into one directory.

Dyspatch variant — connecting your account (optional but recommended). Add the Dyspatch MCP server so the skill can read your modules:

claude mcp add dyspatch -e DYSPATCH_API_KEY=YOUR_KEY -- npx dyspatch-mcp

Without it, the Dyspatch skill still works — it just asks you to paste your module names or falls back to a generic block list.

Usage

Once installed, you don't invoke anything special — just describe what you're doing and the skill triggers itself. For example:

• "Write a brief for my Black Friday email."
• "Help me scope a 3-email welcome series."
• "I need a creative brief for a product launch email."
• "Brief my team on this campaign."

Want to see the format before committing? Ask for a sample and the skill returns an annotated blank template instead of running the interview:

• "Show me what a brief looks like."
• "Give me the brief template."
• "What info do I need to gather for an email brief?"

How the interview flows

1. Picks up where you left off. If you've briefed campaigns before, it reuses what it knows about your brand, ESP, compliance scope, owners, and output format instead of re-asking.
2. Sets the mode. Conversational or batched, single email or sequence, single-locale or multi-locale, and whether light web research is allowed.
3. Gathers the brief, pushing for specifics — a quantitative Goal with a timeframe, one persona, a Single Minded Idea of ≤8 words.
4. Offers recommendations and a sample email (both optional).
5. Shows you the full draft inline and iterates until you're happy — then produces the file.

Repository layout

The skills are plain Markdown; only SKILL.md is always in context, with the rest pulled in on demand. Each skill owns its SKILL.md; the references and assets are single-sourced under shared/ and composed into each skill's package at build time.

Path	Purpose
skills/email-campaign-brief/SKILL.md	Generic skill — entry point and interview flow.
skills/email-campaign-brief-dyspatch/SKILL.md	Dyspatch variant — same flow, Dyspatch as the production path.
skills/email-campaign-brief-dyspatch/references/dyspatch_integration.md	Dyspatch concepts, MCP/API setup, and how the sample layout is grounded in real modules.
shared/references/section_guide.md	What a "good" answer looks like for each section, with examples.
shared/references/email_specifics.md	Detail on the email-only elements (production, localization, UTMs, compliance, …).
shared/references/proven_patterns.md	Curated library of techniques that have worked, with sources.
shared/assets/brief_template.md	The structural template the final brief is assembled from.
shared/assets/sample_brief_template.md	Annotated blank template returned when you ask for a sample.

---

Development

Tooling is plain Node (ESM) with zero npm dependencies, driven by package.json:

Command	What it does
Both skills are built from one source tree (the set is declared in the SKILLS array in scripts/skill.mjs), so every command operates on both.

Command	What it does
npm test	Run the validation suite (node --test) for every skill. Enforces the Agent Skills spec: description ≤ 1024 characters, name format/length, optional compatibility ≤ 500 characters, and that every referenced references//assets/ file exists (skill-local or shared). Also round-trips packaging.
npm run validate	The same checks as a readable ✓/✗ report, grouped per skill.
npm run bump -- patch|minor|major|<x.y.z>	Bump the single shared version in package.json (the source of truth) and sync it into both SKILL.md frontmatters' metadata.version.
npm run package	Build one dist/<name>-<version>.skill per skill (a zip), composing the shared references//assets/ with each skill's own files. Validates first and refuses to package a failing skill.
npm run build	npm test then npm run package.

Packaging requires the zip binary (apt install zip / brew install zip). Each packaged .skill ships only the composed SKILL.md, references/, and assets/ — the tooling (package.json, scripts/, tests/, dist/) and the docs/ marketing site are excluded.