Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Design and write email campaigns and sequences including onboarding flows, lifecycle campaigns, transactional emails, newsletters, broadcast sends, and launch and announcement emails. Use this skill whenever the user wants to write email copy, plan an email sequence, design an onboarding drip, or set up lifecycle email campaigns. Triggers on email sequence, drip campaign, onboarding email, lifecycle email, welcome email, transactional email, newsletter, email broadcast, launch email, announcemen
.claude/skills/rampstackco-email-sequences/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-17 | ✗→✓ | ▲ Improved | 93% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 55% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 69% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 134% | 0% |
Plan and write email campaigns. Sequences (multi-message flows triggered by events) and broadcasts (one-off sends). Stack-agnostic. Works with any email service provider.
landing-page-copy)content-and-copy)brand-voice)analytics-strategy)If the audience is undefined, define it before writing. Generic emails to "everyone" perform worse than targeted ones.
Most email programs run a handful of standard sequence patterns. Each has its own goals, structure, and pitfalls.
Triggered by signup. Goal: get the user to first value.
Typical structure (5 emails over 14 days):
Common failure: Front-loading product features. Users don't care about features yet; they care about getting to value.
Time- or behavior-triggered. Goal: move users from signup to engaged user.
Patterns:
Best practice: Trigger on behavior, not just time. A user who has done 3 onboarding steps doesn't need an email telling them to get started.
Ongoing. Goal: keep active users engaged.
Patterns:
Cadence rule: Frequency that earns the read, not frequency that fills the calendar. Weekly newsletter that is genuinely useful beats daily newsletter that gets archived.
Triggered by inactivity. Goal: pull a lapsed user back.
Typical structure (3 emails):
Best practice: Honor the unsubscribe. Aggressive win-back damages deliverability. A clean list of engaged subscribers beats a large list of inactive ones.
Triggered by actions: receipts, password resets, notifications, order confirmations, shipping updates.
Best practices:
Highest open rates of any email type. Use sparingly for marketing nudges; over-marketing transactional emails damages trust.
One-off sends. Announcements, launches, news, time-sensitive campaigns.
Best practices:
Regardless of sequence type, every email has the same components.
The deciding factor for whether the email gets opened.
Patterns:
Avoid:
Length: 30 to 50 characters. Mobile clients truncate longer.
The line that appears below or beside the subject in most email clients.
Best practice:
The first line of the email body. The reader is deciding whether to keep reading.
Strong openings:
Weak openings:
The substance.
Length guide:
Structure:
The action the email is asking for.
Best practices:
Default output is a markdown document per email or sequence:
markdown# Sequence: [Name] **Trigger:** [What starts this sequence] **Goal:** [Outcome at completion] **Audience:** [Specific segment] **Length:** [N emails over X days] ## Email 1 - [Subject working title] **Send:** [Trigger / Day N] **Subject:** [text] **Preview:** [text] [Body] **Primary CTA:** [text] **CTA URL:** [destination] --- ## Email 2 - [Subject working title] [Same structure] --- [etc.]
For a single broadcast, just one email block.
references/subject-line-patterns.md - Subject line patterns with examples for each sequence type.references/sequence-templates.md - Skeleton templates for the 6 sequence types.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | fail→pass | 13,568 | 12,926 | -5% | 1 | 1 | 0% | 2,342 | 4,525 | +93% | 0 | 0 | — |
case-18 | pass→pass | 8,642 | 6,563 | -24% | 1 | 1 | 0% | 1,419 | 3,433 | +142% | 0 | 0 | — |
case-01 | fail→pass | 14,253 | 10,438 | -27% | 1 | 1 | 0% | 2,619 | 4,054 | +55% | 0 | 0 | — |
case-02 | fail→fail | 11,378 | 7,782 | -32% | 1 | 1 | 0% | 1,825 | 3,557 | +95% | 0 | 0 | — |
case-03 | fail→fail | 18,977 | 14,076 | -26% | 1 | 1 | 0% | 3,024 | 4,573 | +51% | 0 | 0 | — |
case-04 | fail→pass | 20,397 | 10,821 | -47% | 1 | 1 | 0% | 3,142 | 4,227 | +35% | 0 | 0 | — |
case-05 | pass→pass | 5,214 | 8,885 | +70% | 1 | 1 | 0% | 903 | 3,859 | +327% | 0 | 0 | — |
case-10 | pass→pass | 6,468 | 4,730 | -27% | 1 | 1 | 0% | 1,052 | 3,142 | +199% | 0 | 0 | — |
case-06 | pass→fail | 7,421 | 7,597 | +2% | 1 | 1 | 0% | 1,242 | 3,647 | +194% | 0 | 0 | — |
case-07 | pass→pass | 9,396 | 10,095 | +7% | 1 | 1 | 0% | 1,439 | 3,991 | +177% | 0 | 0 | — |
case-08 | pass→pass | 4,105 | 5,922 | +44% | 1 | 1 | 0% | 659 | 3,302 | +401% | 0 | 0 | — |
case-09 | pass→pass | 11,768 | 11,904 | +1% | 1 | 1 | 0% | 1,836 | 4,111 | +124% | 0 | 0 | — |
case-16 | fail→pass | 14,317 | 9,774 | -32% | 1 | 1 | 0% | 2,346 | 3,975 | +69% | 0 | 0 | — |
case-11 | pass→pass | 11,336 | 9,810 | -13% | 1 | 1 | 0% | 1,704 | 3,898 | +129% | 0 | 0 | — |
case-12 | fail→pass | 9,883 | 7,710 | -22% | 1 | 1 | 0% | 1,522 | 3,555 | +134% | 0 | 0 | — |
case-13 | pass→pass | 16,349 | 8,299 | -49% | 1 | 1 | 0% | 2,426 | 3,578 | +47% | 0 | 0 | — |
case-14 | pass→pass | 14,951 | 14,301 | -4% | 1 | 1 | 0% | 2,159 | 4,479 | +107% | 0 | 0 | — |
case-15 | pass→pass | 11,867 | 11,376 | -4% | 1 | 1 | 0% | 1,807 | 4,108 | +127% | 0 | 0 | — |
case-19 | pass→pass | 12,576 | 8,568 | -32% | 1 | 1 | 0% | 1,835 | 3,674 | +100% | 0 | 0 | — |
case-20 | pass→pass | 14,189 | 11,755 | -17% | 1 | 1 | 0% | 2,150 | 4,234 | +97% | 0 | 0 | — |
case-21 | pass→pass | 17,559 | 17,355 | -1% | 1 | 1 | 0% | 2,789 | 5,041 | +81% | 0 | 0 | — |
case-22 | pass→pass | 16,671 | 21,658 | +30% | 1 | 1 | 0% | 2,925 | 6,299 | +115% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +18 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.