Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Viet brief anh static day du cho designer hoac photographer: muc tieu visual, concept, mood, composition, copy overlay, mau va font doc tu brand guideline, kich thuoc theo kenh, tieu chi duyet. Kich hoat khi nhac: brief anh, brief visual, brief banner, brief thumbnail, brief static, brief hinh anh.
.claude/skills/minhnv0807-42-brief-hinh-anh/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 86% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 91% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 105% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 72% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 67% | 0% |
> Brief anh tot = designer lam xong ma khong can mot cuoc goi nao. Truyen dat cam xuc muon co, khong chi "can anh dep". Mau, font, logo doc tu brand guideline (46-brand-guideline) — khong tu bia.
Doc .agents/product-marketing-context.md, brand guideline va asset list (41-campaign-asset-list) neu co. Neu thieu, hoi toi da 4 cau:
50-asset-resize).Truoc khi brief concept, xac dinh anh nay can lam duoc gi trong 3 giay dau nguoi xem nhin thay:
| Muc tieu | Huong visual | |----------|--------------| | Stop scroll | Hook manh, tuong phan cao, pattern interrupt | | Build trust | Clean, professional, social proof nhin thay duoc | | Drive click | CTA noi bat, offer ro, urgency neu co |
Dien het cac section trong template ben duoi — khong de trong section nao.
Luon kem 2-3 anh reference cu the, ghi ro hoc gi tu moi reference (mau? composition? mood?) va dieu khong copy.
Ten file: brief-hinh-anh-[ten-asset]-[YYYYMMDD].md
markdown# Brief Anh — [Ten asset] ## Thong tin chung | Hang muc | Chi tiet | |----------|----------| | Ten project / campaign | | | Loai anh | | | Kenh / vi tri dang | | | Kich thuoc (px) + ti le | | | Muc tieu visual | Stop scroll / Build trust / Drive click | | Format xuat | JPG / PNG / WebP | | So luong deliverables | | | Designer | | | Deadline | | | So lan sua toi da | | ## Concept & Mood - Idea mot cau: [anh nay noi ve dieu gi — doc xong la hinh dung duoc] - Cam xuc muon truyen: [Professional / Warm / Playful / Urgent / Minimal / Bold] - Mood board reference: - Ref 1: [link/file — hoc gi: mau? layout? kieu anh?] - Ref 2: [...] - Ref 3: [...] — khong copy: [...] - Mau chu dao (doc tu brand guideline): - Primary: [mau primary] - Secondary / accent: [mau secondary] - Background: [trang / mau solid / gradient / anh that] - Text: [mau text dam tren nen sang / trang tren nen toi] - Tuyet doi tranh: mau [...], style [...], element [...] ## Composition - Subject chinh: [Nguoi that / San pham / Abstract / Text-based / Illustration] - Background: [Trang / Mau solid / Lifestyle scene / Studio / Blur / Texture] - Focal point: [Giua khung / Rule of thirds trai / phai / Full bleed] - Khoang trong cho text overlay: [co/khong — goc nao, chiem khoang 1/3 nao] - Hierarchy visual: 1. [Element quan trong nhat — mat nhin vao dau tien] 2. [Element thu hai] 3. [Element thu ba — CTA / logo] ## Copy overlay (da duyet truoc khi embed) | Element | Noi dung | Font | Size | Mau | Vi tri | |---------|----------|------|------|-----|--------| | Headline | | [font heading] Bold | | | | | Subtext | | [font body] Regular | | | | | CTA button | | [font heading] Bold | | | | | Logo | [version: ngang / dung / icon only] | — | — | — | [theo brand guideline] | Luu y copy: headline toi da [X] chu, subtext toi da [X] chu, khong qua 3 mau text trong 1 anh. ## Deliverables - [ ] File goc (share link Canva/Figma — khong export flatten) - [ ] Export [format] — [kich thuoc cu the] - [ ] Variants neu can: [light/dark version, crop khac] - [ ] Resize cac kich thuoc con lai theo asset list ## Quy trinh review - Nop draft 1: [ngay] — feedback trong [X] gio - Sua toi da: [X] lan — approve cuoi boi: [Leader] - Deadline final: [ngay]
| Kenh | Kich thuoc | Safe zone | Format export | Max file size | |------|-----------|-----------|---------------|---------------| | Facebook/IG Feed | 1080x1080px (1:1) hoac 1080x1350px (4:5) | 20px moi canh | JPG, RGB | 30MB | | Facebook/IG Story | 1080x1920px (9:16) | 250px top / 250px bottom | JPG, RGB | 30MB | | TikTok thumbnail | 1080x1920px (9:16) | 150px top / 300px bottom | JPG | 10MB | | Email header | 600x200px | 10px moi canh | JPG | 1MB | | Website banner | 1440x500px | 100px moi canh | JPG/PNG | 500KB |
Bang kich thuoc day du cac kenh: xem 50-asset-resize.
Brief anh — Facebook Feed 1080x1080 — TOFU campaign khoa hoc Q3
Loai: single image ads | Kenh: Facebook Feed | Muc tieu: stop scroll trong 1.5 giay dau
Cam xuc: bold, urgency nhe, professional nhung khong corporate
Subject: text-based, background gradient [mau primary] sang [mau secondary]
Headline: "[thong diep chinh — toi da 60 ky tu]"
Subtext: "[ten san pham + moc thoi gian]"
CTA: "Dang ky ngay"
Ref: [link anh text-based bold, hoc contrast va typography, khong copy mau]
Deadline: [ngay] | Sua toi da: 2 lan36-content-brief: content brief tong the — brief anh la buoc chi tiet hoa phan visual.41-campaign-asset-list: xac dinh anh nay nam o dau trong danh sach asset + deadline.46-brand-guideline: nguon mau, font, logo rules — doc truoc khi viet brief.45-brief-canva: khi designer lam tren Canva, kem direction thao tac cu the.47-design-review: review ket qua theo tieu chi cham diem sau khi designer nop.30-thiet-ke-master: khi can gen anh bang AI thay vi designer lam tay.74-copy-hai-lop: khi poster can cau hai lop / slogan — chot cau o 74 roi moi brief anh; 42 khong tu viet slogan.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 22,119 | 29,627 | +34% | 1 | 1 | 0% | 3,637 | 6,257 | +72% | 0 | 0 | — |
case-02 | fail→pass | 18,856 | 27,637 | +47% | 1 | 1 | 0% | 3,373 | 6,281 | +86% | 0 | 0 | — |
case-03 | fail→fail | 21,900 | 23,626 | +8% | 1 | 1 | 0% | 3,733 | 6,220 | +67% | 0 | 0 | — |
case-04 | fail→fail | 15,723 | 16,640 | +6% | 1 | 1 | 0% | 2,811 | 4,988 | +77% | 0 | 0 | — |
case-05 | fail→fail | 21,539 | 19,361 | -10% | 1 | 1 | 0% | 3,542 | 5,738 | +62% | 0 | 0 | — |
case-06 | pass→pass | 11,884 | 10,316 | -13% | 1 | 1 | 0% | 2,060 | 4,064 | +97% | 0 | 0 | — |
case-07 | fail→pass | 17,101 | 19,480 | +14% | 1 | 1 | 0% | 2,840 | 5,435 | +91% | 0 | 0 | — |
case-08 | fail→fail | 15,750 | 20,228 | +28% | 1 | 1 | 0% | 2,604 | 5,616 | +116% | 0 | 0 | — |
case-09 | fail→pass | 15,240 | 17,874 | +17% | 1 | 1 | 0% | 2,394 | 4,907 | +105% | 0 | 0 | — |
case-10 | pass→pass | 15,918 | 11,823 | -26% | 1 | 1 | 0% | 2,771 | 4,207 | +52% | 0 | 0 | — |
case-11 | fail→pass | 10,916 | 6,330 | -42% | 1 | 1 | 0% | 1,978 | 3,399 | +72% | 0 | 0 | — |
case-12 | fail→fail | 13,961 | 18,848 | +35% | 1 | 1 | 0% | 2,408 | 5,185 | +115% | 0 | 0 | — |
case-13 | fail→fail | 16,171 | 13,251 | -18% | 1 | 1 | 0% | 2,606 | 4,324 | +66% | 0 | 0 | — |
case-14 | pass→pass | 9,455 | 5,363 | -43% | 1 | 1 | 0% | 1,791 | 3,287 | +84% | 0 | 0 | — |
case-15 | pass→pass | 16,078 | 20,949 | +30% | 1 | 1 | 0% | 2,674 | 5,702 | +113% | 0 | 0 | — |
case-16 | fail→pass | 18,089 | 15,861 | -12% | 1 | 1 | 0% | 2,870 | 4,788 | +67% | 0 | 0 | — |
case-17 | fail→pass | 15,839 | 13,014 | -18% | 1 | 1 | 0% | 2,553 | 4,301 | +68% | 0 | 0 | — |
case-18 | fail→pass | 12,836 | 8,850 | -31% | 1 | 1 | 0% | 2,201 | 3,750 | +70% | 0 | 0 | — |
case-19 | pass→pass | 15,181 | 16,993 | +12% | 1 | 1 | 0% | 2,639 | 4,967 | +88% | 0 | 0 | — |
case-20 | pass→pass | 11,982 | 4,734 | -60% | 1 | 1 | 0% | 2,027 | 2,996 | +48% | 0 | 0 | — |
case-21 | pass→pass | 14,655 | 14,843 | +1% | 1 | 1 | 0% | 2,513 | 4,654 | +85% | 0 | 0 | — |
case-22 | pass→pass | 14,904 | 10,604 | -29% | 1 | 1 | 0% | 2,684 | 3,907 | +46% | 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 +32 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are 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.