← All posts

Email Marketing Is Three Skills, and the Copy Is the Easy One

2026-11-10

Most teams treat email marketing as a writing task with some formatting at the end. The scores in our directory say otherwise: the writing skill is the least stressed link in the chain, and the skill that touches the actual inbox is pure infrastructure.


A pipeline, not a task

The three email-marketing skills in our directory cover one workflow, split exactly where the work changes character. email-creative-builder — 9.0 overall — drafts the copy: subject lines, body variants, the creative itself. email-render-builder — 9.0 — turns that copy into responsive HTML, and explicitly handles dark-mode rendering, which is where most hand-built emails quietly break. deliverability-qa — 9.0 — runs the pre-flight: SPF, DKIM, DMARC, and everything else that decides whether the message arrives at all.

📣 email-creative-builder 9.0 Use when the user asks to "write the email", "draft subject lines", or "build email creative"; pr… ⬇ 11 📣 email-render-builder 9.0 Use when the user asks to "build the email HTML", "make this email responsive", "fix dark-mode re… ⬇ 13 📣 deliverability-qa 9.0 Use when the user asks to "run a deliverability pre-flight before I send", "check my SPF/DKIM/DMA… ⬇ 14

The split matters because each stage fails differently. Copy fails persuasively — you can read it and see why it is bad. HTML fails visually — you open it in Outlook and the layout collapses. Deliverability fails silently: the email "sends," the bounce report stays clean, and the message is in spam or nowhere. The failure modes get less visible as you move down the pipeline, which is precisely why the later stages need tooling more than the earlier ones.

Where the points come from

All three score 9.5 on trigger quality, and they earn it: each description names the exact user phrasings it should fire on — "write the email," "build the email HTML," "run a deliverability pre-flight before I send." There is no ambiguity about which skill owns which sentence. That is rarer than it sounds. In most multi-skill families we review, at least two members reach for the same prompt.

Workflow scores are 9.2 for the creative skill and 9.6 for both the render and QA skills — the rendering and QA skills are more procedural, and procedural is what the workflow dimension rewards. Content sits at 9.0 across the board. Security is 9.5 for all three, which mostly reflects that none of them do anything reckless with credentials or external services.

The shared soft spot is engineering: 7 for the creative builder, 6 for the other two. The skills tell the agent what to check and in what order, but they lean on prose where a validation script would be stronger. For a deliverability skill especially, you want check-spf --domain example.com returning a pass/fail, not a paragraph about why SPF matters.

Deliverability is an engineering problem

The interesting claim in this family is buried in the QA skill's framing: deliverability is infrastructure, not copywriting. Open rates get attributed to subject lines because subject lines are what marketers can see and edit. But a failed DMARC alignment or a missing unsubscribe header can zero out an entire campaign before anyone reads a word, and no amount of copy revision fixes it. The skill treats authentication records, list hygiene, and sending reputation as pre-send gates — the same way you treat a failing CI check before a deploy.

Downloads tell the same story from the demand side: 11 for the creative skill, 13 for the render builder, 14 for the QA skill. The least glamorous skill in the chain is the most installed. People who actually send email discover quickly that the inbox is where their problems live.

The ceiling

None of the three breaks 9.1, and the reason is consistent: they are advice-shaped rather than tool-shaped. The descriptions and procedures are excellent — trigger quality at 9.5 is close to the ceiling of our directory — but the engineering dimension at 6-7 means the agent is left to interpret instructions that a script could verify deterministically. The natural next version of this pipeline is one where the QA stage runs executable checks and refuses to pass the campaign, the same way a build gate refuses to pass a broken commit. Until then, the skills are the best description of the work — and the work is engineering.