Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Code email HTML responsive dung duoc tren moi email client: cau truc table-based, inline CSS, max 600px, dark mode, spec tung client (Gmail, Outlook, Apple Mail), checklist test truoc khi gui. Mau va font doc tu brand guideline. Kich hoat khi nhac: email html, code email, template email html, email responsive.
.claude/skills/minhnv0807-49-html-email-template/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 164% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 79% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 69% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 137% | 0% |
> Email HTML khac website HTML: phai dung table layout, inline CSS, tranh moi thu client khong support. Tieu chi: dep tren Gmail, Outlook, Apple Mail, doc duoc tren mobile 375px — khong co exception. Noi dung va sequence lay tu 14-email-marketing; skill nay lo phan code.
Doc brand guideline (46-brand-guideline) va noi dung email tu 14-email-marketing neu co. Neu thieu, hoi toi da 4 cau:
<style> (chi giu <style> cho media query mobile va dark mode, phan con lai inline het).| Layout | Dung cho | |--------|----------| | Simple 1 cot | Broadcast, nurture — de doc nhat, it loi nhat | | 2 cot (hinh + text) | Offer email co hinh san pham | | Newsletter nhieu muc | Digest, update — moi muc 1 block table rieng |
Container: 600px desktop / 100% mobile
Background ngoai: mau xam nhat (vd #F4F4F4)
Background body: [mau nen] (thuong trang)
Font: '[font body]', Arial, Helvetica, sans-serif
Body font-size: 16px / line-height: 1.6
Mobile breakpoint: max-width 480pxDesign tokens doc tu brand guideline:
[dia chi], unsubscribe "Ban nhan email nay vi da dang ky tai website] | Huy dang ky", copyright © nam] ten thuong hieu].css@media only screen and (max-width: 480px) { /* container: 100% */ /* body font: 15px */ /* CTA: width 100%, can giua */ /* padding: 24px 20px */ /* image: 100% width */ }
Dark mode:
<meta name="color-scheme" content="light dark"> va <meta name="supported-color-schemes" content="light dark">.@media (prefers-color-scheme: dark) chi duoc ho tro mot phan: Apple Mail ho tro tot; Gmail thuong tu dao mau va bo qua media query; Outlook co the tu invert.{{first_name}} — ten nguoi nhan (doi cu phap theo tool gui){{unsubscribe_link}} — bat buoc phap ly[dia chi cong ty] — dien that[CTA_LINK] — dien link that (kem UTM theo 14-email-marketing)| Client | Dac diem render | Luu y khi code | |--------|-----------------|----------------| | Gmail (web/app) | Co the strip <style> trong mot so ngu canh | Inline toan bo CSS; media query co the bi bo qua — layout 1 cot van phai doc duoc | | Outlook desktop (Windows) | Render bang Word engine | Chi table; khong border-radius, khong background-image day du (dung mau solid thay the); test rieng | | Apple Mail / iPhone Mail | WebKit, render tot nhat | Ho tro media query + dark mode | | Yahoo / webmail khac | Ho tro kha, khac nhau tung ban | Gui test that de kiem tra | | Mobile apps noi chung | Man hinh 375px | Font >= 15px, CTA cao toi thieu 44px de bam bang ngon tay |
14-email-marketing: chien luoc, sequence, noi dung, subject line — viet truoc, code sau.46-brand-guideline: mau, font, logo cho email tokens.47-design-review: review email truoc khi gui (brand + readability + CTA + platform fit).12-brief-landing-page: CTA email thuong tro ve landing page — dong bo message giua 2 ben.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 24,133 | 38,269 | +59% | 1 | 1 | 0% | 5,262 | 10,424 | +98% | 0 | 0 | — |
case-02 | fail→fail | 20,642 | 32,995 | +60% | 1 | 1 | 0% | 4,826 | 9,779 | +103% | 0 | 0 | — |
case-03 | fail→pass | 30,734 | 33,505 | +9% | 1 | 1 | 0% | 7,063 | 9,943 | +41% | 0 | 0 | — |
case-04 | fail→fail | 27,966 | 40,072 | +43% | 1 | 1 | 0% | 4,554 | 10,378 | +128% | 0 | 0 | — |
case-05 | fail→fail | 34,350 | 35,923 | +5% | 1 | 1 | 0% | 5,968 | 8,511 | +43% | 0 | 0 | — |
case-06 | fail→fail | 24,891 | 26,524 | +7% | 1 | 1 | 0% | 4,444 | 6,789 | +53% | 0 | 0 | — |
case-07 | fail→pass | 13,468 | 24,078 | +79% | 1 | 1 | 0% | 2,827 | 7,452 | +164% | 0 | 0 | — |
case-08 | fail→fail | 17,131 | 24,061 | +40% | 1 | 1 | 0% | 3,655 | 7,231 | +98% | 0 | 0 | — |
case-09 | fail→pass | 23,007 | 30,235 | +31% | 1 | 1 | 0% | 4,778 | 8,551 | +79% | 0 | 0 | — |
case-10 | pass→pass | 13,451 | 16,990 | +26% | 1 | 1 | 0% | 2,919 | 5,920 | +103% | 0 | 0 | — |
case-11 | pass→pass | 12,124 | 13,200 | +9% | 1 | 1 | 0% | 2,328 | 4,585 | +97% | 0 | 0 | — |
case-12 | fail→pass | 14,514 | 12,945 | -11% | 1 | 1 | 0% | 2,852 | 4,813 | +69% | 0 | 0 | — |
case-13 | pass→fail | 16,691 | 15,597 | -7% | 1 | 1 | 0% | 3,312 | 5,125 | +55% | 0 | 0 | — |
case-14 | pass→pass | 17,112 | 21,359 | +25% | 1 | 1 | 0% | 3,458 | 6,799 | +97% | 0 | 0 | — |
case-15 | pass→pass | 13,300 | 17,264 | +30% | 1 | 1 | 0% | 2,483 | 5,641 | +127% | 0 | 0 | — |
case-16 | fail→pass | 15,142 | 38,131 | +152% | 1 | 1 | 0% | 2,413 | 5,716 | +137% | 0 | 0 | — |
case-17 | pass→pass | 11,057 | 11,015 | -0% | 1 | 1 | 0% | 1,993 | 3,985 | +100% | 0 | 0 | — |
case-18 | pass→pass | 18,817 | 17,582 | -7% | 1 | 1 | 0% | 3,427 | 5,587 | +63% | 0 | 0 | — |
case-19 | pass→pass | 18,267 | 20,970 | +15% | 1 | 1 | 0% | 3,143 | 5,191 | +65% | 0 | 0 | — |
case-20 | pass→pass | 15,515 | 16,621 | +7% | 1 | 1 | 0% | 2,673 | 4,586 | +72% | 0 | 0 | — |
case-21 | fail→fail | 15,949 | 16,454 | +3% | 1 | 1 | 0% | 3,021 | 5,413 | +79% | 0 | 0 | — |
case-22 | pass→pass | 10,109 | 8,879 | -12% | 1 | 1 | 0% | 1,772 | 3,706 | +109% | 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.