Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Email delivery infrastructure - DNS authentication (SPF/DKIM/DMARC), subdomain isolation, provider abstraction, template systems, bounce handling, warmup strategy, and deliverability monitoring.
.claude/skills/email-infrastructure/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | — | — |
| case-03 | ✗→✓ | ▲ Improved | — | — |
| case-11 | ✗→✓ | ▲ Improved | — | — |
| case-05 | ✗→✓ | ▲ Improved | — | — |
| case-12 | ✗→✓ | ▲ Improved | — | — |
Production email delivery requires DNS authentication, domain isolation, and provider-agnostic architecture. A single misconfiguration can land your entire domain in spam.
BAD: Send bulk marketing from example.com
→ Spam complaints tank your main domain reputation
→ Transactional emails (password reset, receipts) start landing in spam
→ Recovery takes weeks of warmup
GOOD: Subdomain isolation with separate reputations
→ mail.example.com for transactional (password reset, receipts, 2FA)
→ notify.example.com for product notifications (comments, mentions)
→ marketing.example.com for bulk campaigns (newsletters, promotions)
→ Each subdomain has independent reputation — one bad campaign does not poison the rest# SPF — declare which servers can send from your domain
mail.example.com TXT "v=spf1 include:_spf.provider.com ~all"
# DKIM — cryptographic signature on every email
selector._domainkey.mail.example.com TXT "v=DKIM1; k=rsa; p=MIGf..."
# DMARC — policy for failed authentication (progressive rollout)
# Week 1-2: monitor only
_dmarc.mail.example.com TXT "v=DMARC1; p=none; rua=mailto:dmarc@example.com"
# Week 3-4: quarantine suspicious emails
_dmarc.mail.example.com TXT "v=DMARC1; p=quarantine; pct=25; rua=mailto:dmarc@example.com"
# Week 5+: reject after confidence builds
_dmarc.mail.example.com TXT "v=DMARC1; p=reject; rua=mailto:dmarc@example.com"Never jump straight to p=reject. The progressive rollout catches misconfigurations before they block legitimate mail.
typescript// Swap Resend, SES, Postmark, or Mailgun without touching business logic interface EmailProvider { send(message: EmailMessage): Promise<EmailResult> sendBatch(messages: EmailMessage[]): Promise<EmailResult[]> } interface EmailMessage { from: string to: string | string[] subject: string html: string text?: string replyTo?: string headers?: Record<string, string> tags?: Record<string, string> } interface EmailResult { id: string status: 'sent' | 'queued' | 'failed' error?: string } // Factory selects provider from config — no hardcoded vendor function createEmailProvider(config: { provider: string }): EmailProvider { switch (config.provider) { case 'resend': return new ResendProvider() case 'ses': return new SESProvider() case 'postmark': return new PostmarkProvider() default: throw new Error(`Unknown email provider: ${config.provider}`) } }
typescriptinterface EmailService { sendTransactional(message: EmailMessage): Promise<EmailResult> sendMarketing(message: EmailMessage): Promise<EmailResult> } class ProductionEmailService implements EmailService { constructor( private transactional: EmailProvider, // high-deliverability provider private marketing: EmailProvider // bulk-optimized provider ) {} async sendTransactional(message: EmailMessage): Promise<EmailResult> { // Transactional: password reset, receipts, 2FA — must arrive instantly // Use mail.example.com subdomain, high-priority provider return this.transactional.send({ ...message, from: `noreply@mail.example.com`, headers: { 'X-Priority': '1' } }) } async sendMarketing(message: EmailMessage): Promise<EmailResult> { // Marketing: newsletters, promotions — rate-limited, includes unsubscribe // Use marketing.example.com subdomain, bulk provider return this.marketing.send({ ...message, from: `hello@marketing.example.com`, headers: { 'List-Unsubscribe': `<https://example.com/unsubscribe>` } }) } }
typescript// MJML compiles to responsive HTML that works across all email clients // Compile at build time, not runtime import mjml2html from 'mjml' const mjmlTemplate = ` <mjml> <mj-head> <mj-attributes> <mj-all font-family="system-ui, -apple-system, sans-serif" /> <mj-text font-size="16px" line-height="1.5" color="#1a1a1a" /> </mj-attributes> <mj-style> @media (prefers-color-scheme: dark) { .dark-bg { background-color: #1a1a1a !important; } .dark-text { color: #e5e5e5 !important; } } </mj-style> </mj-head> <mj-body> <mj-section css-class="dark-bg"> <mj-column> <mj-text css-class="dark-text">Hello {{name}},</mj-text> <mj-text css-class="dark-text">{{body}}</mj-text> <mj-button href="{{actionUrl}}" background-color="#2563eb"> {{actionLabel}} </mj-button> </mj-column> </mj-section> </mj-body> </mjml> ` function escapeHtml(str: string): string { return str.replace(/&/g, '&').replace(/</g, '<').replace(/>/g, '>').replace(/"/g, '"') } function compileTemplate(mjml: string, vars: Record<string, string>): string { let compiled = mjml for (const [key, value] of Object.entries(vars)) { compiled = compiled.replaceAll(`{{${key}}}`, escapeHtml(value)) } const { html, errors } = mjml2html(compiled) if (errors.length > 0) { throw new Error(`MJML compilation errors: ${errors.map(e => e.message).join(', ')}`) } return html }
typescript// Webhook handler for provider callbacks (bounces, complaints, deliveries) interface BounceEvent { type: 'bounce' | 'complaint' | 'delivery' email: string reason?: string timestamp: string } async function handleEmailWebhook(event: BounceEvent): Promise<void> { switch (event.type) { case 'bounce': // Hard bounce: address does not exist — never send again await db.emailSuppression.upsert({ where: { email: event.email }, create: { email: event.email, reason: 'hard_bounce', suppressedAt: new Date() }, update: { reason: 'hard_bounce', suppressedAt: new Date() } }) break case 'complaint': // Spam complaint: user marked as spam — suppress immediately await db.emailSuppression.upsert({ where: { email: event.email }, create: { email: event.email, reason: 'complaint', suppressedAt: new Date() }, update: { reason: 'complaint', suppressedAt: new Date() } }) // Alert if complaint rate exceeds 0.1% (ISP threshold) await checkComplaintRate() break case 'delivery': await db.emailLog.update({ where: { email: event.email }, data: { deliveredAt: new Date() } }) break } } // Always check suppression list before sending async function isSuppressed(email: string): Promise<boolean> { const entry = await db.emailSuppression.findUnique({ where: { email } }) return entry !== null }
New domain/IP starts with zero reputation. Send too fast and ISPs block you.
Week 1: 50 emails/day → Send to your most engaged users only
Week 2: 200 emails/day → Expand to users who opened in last 30 days
Week 3: 500 emails/day → Include 90-day active users
Week 4: 1,000 emails/day → General audience, monitor bounce rate
Week 5: 5,000 emails/day → Scale up if bounce < 2% and complaints < 0.1%
Week 6: 10,000 emails/day → Full volume if metrics stay clean
Week 8+: Full send → Maintain list hygiene going forward
CRITICAL THRESHOLDS:
Bounce rate > 5% → STOP sending, clean your list
Complaint rate > 0.1% → STOP sending, review content and targeting
Open rate < 10% → Re-evaluate subject lines and audiencetypescriptinterface DeliverabilityMetrics { sent: number delivered: number bounced: number complained: number opened: number clicked: number } async function getDailyMetrics(date: string): Promise<DeliverabilityMetrics> { const metrics = await db.emailLog.aggregate({ where: { sentAt: { gte: new Date(date), lt: new Date(date + 'T23:59:59Z') } }, _count: { id: true }, // Group by status for each metric }) return metrics } async function checkHealthThresholds(metrics: DeliverabilityMetrics): Promise<void> { if (metrics.sent === 0) return const bounceRate = metrics.bounced / metrics.sent const complaintRate = metrics.complained / metrics.sent const deliveryRate = metrics.delivered / metrics.sent if (bounceRate > 0.05) { await alertOps('Bounce rate critical', `${(bounceRate * 100).toFixed(1)}% — pause sending`) } if (complaintRate > 0.001) { await alertOps('Complaint rate critical', `${(complaintRate * 100).toFixed(2)}% — review content`) } if (deliveryRate < 0.95) { await alertOps('Delivery rate low', `${(deliveryRate * 100).toFixed(1)}% — check DNS and reputation`) } }
Before every send, verify:
1. Recipient is not on suppression list (bounces + complaints)
2. SPF/DKIM/DMARC records are valid for the sending subdomain
3. Unsubscribe link is present (CAN-SPAM, GDPR requirement)
4. Plain text version exists alongside HTML
5. From address matches the authenticated subdomain
6. List-Unsubscribe header is set for bulk sends
7. Subject line is not empty and under 78 charactersKey principle: Treat email infrastructure like a reputation system. Subdomain isolation protects your core domain. Progressive DMARC rollout catches issues before they block mail. Always check the suppression list before sending. Monitor bounce and complaint rates daily — by the time you notice spam folder placement, the damage is already done.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-03 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-02 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-10 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-13 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-09 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-11 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-15 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-18 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-17 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-08 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-07 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-21 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-23 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-19 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-14 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-01 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-05 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-06 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-12 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-16 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-20 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-22 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
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. 23 cases were attempted, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +26 percentage points is the difference between those two pass rates over the 22 comparable cases.
The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.
Other measured skills in the registry, with their headline benchmark lift.