Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Leader review / duyet output cua team truoc khi trien khai: checklist duyet content brief (angle dung insight, dung pillar, CTA ro), checklist duyet ads copy + creative (claim co proof, dung tone, hook trong 125 ky tu dau), verdict Approve / Revise / Redo kem feedback template. Dung khi nhac 'duyet brief', 'review ads', 'duyet content', 'approve creative'.
.claude/skills/minhnv0807-62-marketing-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 53% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 61% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 147% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 79% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 92% | 0% |
> Review brief truoc khi viet = sua 1 lan. Khong review = sua 5 lan sau khi viet xong. Review ads truoc khi launch = ngan lang phi budget vao creative sai; 1 gio review tiet kiem nhieu ngay CPL te. Leader review khong phai de "thich hay khong thich" — ma de dam bao dung objective, dung audience, dung brand. Visual/design → dung 47-design-review.
Hoi toi da 4 cau:
| Loai output | Dung checklist | Ghi chu | |-------------|----------------|---------| | Content brief (truoc khi writer viet) | Mode A ben duoi | | | Ads copy + creative (truoc khi bat ads) | Mode B ben duoi | | | Visual / design (banner, carousel, LP) | 47-design-review | Skill rieng cho visual |
| Hang muc | Danh gia | Ghi chu | |----------|----------|---------| | Du thong tin de lam viec chua? | Du / Thieu | | | Dung tang pheu chua? | OK / Warning / Fail | | | Dung audience + insight chua? | OK / Warning / Fail | | | Message nhat quan voi campaign? | OK / Warning / Fail | | | Brand-safe (claim, legal, policy)? | OK / Fail | | | Verdict | Approve / Can sua / Redo | |
Objective & tang pheu:
Audience & insight:
Angle & hook:
Message & CTA:
Format & proof:
Brand & legal (fail 1 muc = khong duyet):
Hook (quyet dinh 80% hieu qua):
Copy structure:
Funnel check:
Ky thuat:
Diem manh cua output nay: ghi 1-2 diem — de nguoi lam biet giu gi]
| # | Van de | Tai sao can sua | Goi y sua the nao | Muc do | |---|--------|-----------------|-------------------|--------| | 1 | | | | Blocker / Nen sua | | 2 | | | | |
| Verdict | Y nghia | Next step | |---------|---------|-----------| | APPROVE | Dat chuan | Writer viet ngay / Media buyer launch trong 24h | | MINOR REVISION | Sua X] diem cu the | Sua xong khong can duyet lai — tu launch | | REDO / REBRIEF | Sai huong, can lam lai | Ghi ro ly do; neu la brief → viet lai brief; Leader + nguoi lam ngoi cung 15 phut |
Ten file: marketing-review-[ten-hang-muc]-[YYYYMMDD].md
markdown# Review — [Content brief / Ads] — [Ten campaign] Nguoi review: [Leader] | Ngay: [date] | Lan review: [1/2] ## 1. Quick verdict | Hang muc | Danh gia | Ghi chu | ## 2. Checklist chi tiet [Mode A hoac Mode B — tick tung muc, ghi chu muc fail] ## 3. Feedback - Diem manh: | # | Van de | Tai sao | Goi y sua | Muc do | ## 4. Verdict: APPROVE / MINOR REVISION / REDO - Next step + deadline:
36-content-brief: brief content duoc tao tu skill nay la doi tuong review Mode A.05-copy-quang-cao: ads copy duoc viet tu skill nay la doi tuong review Mode B.47-design-review: visual/design co checklist rieng — khong duyet visual bang skill nay.02-brief-chien-dich: campaign brief goc la chuan de doi chieu message nhat quan.64-team-brief: khi REDO, dung team-brief de giao lai task voi definition of done ro hon.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | pass→pass | 18,475 | 22,717 | +23% | 1 | 1 | 0% | 2,793 | 4,905 | +76% | 0 | 0 | — |
case-10 | pass→pass | 12,320 | 12,134 | -2% | 1 | 1 | 0% | 1,993 | 4,069 | +104% | 0 | 0 | — |
case-01 | fail→pass | 24,778 | 23,013 | -7% | 1 | 1 | 0% | 3,401 | 5,191 | +53% | 0 | 0 | — |
case-02 | fail→pass | 25,450 | 24,660 | -3% | 1 | 1 | 0% | 3,359 | 5,411 | +61% | 0 | 0 | — |
case-03 | fail→fail | 24,699 | 27,996 | +13% | 1 | 1 | 0% | 3,194 | 5,775 | +81% | 0 | 0 | — |
case-04 | fail→pass | 15,922 | 18,422 | +16% | 1 | 1 | 0% | 1,737 | 4,288 | +147% | 0 | 0 | — |
case-05 | fail→fail | 24,308 | 24,737 | +2% | 1 | 1 | 0% | 2,456 | 5,232 | +113% | 0 | 0 | — |
case-06 | fail→fail | 27,223 | 31,966 | +17% | 1 | 1 | 0% | 3,893 | 6,678 | +72% | 0 | 0 | — |
case-07 | pass→pass | 21,460 | 23,654 | +10% | 1 | 1 | 0% | 2,734 | 5,395 | +97% | 0 | 0 | — |
case-08 | pass→pass | 18,145 | 16,737 | -8% | 1 | 1 | 0% | 2,223 | 4,065 | +83% | 0 | 0 | — |
case-11 | pass→pass | 17,264 | 18,653 | +8% | 1 | 1 | 0% | 2,123 | 4,424 | +108% | 0 | 0 | — |
case-12 | pass→pass | 14,873 | 19,755 | +33% | 1 | 1 | 0% | 2,535 | 4,614 | +82% | 0 | 0 | — |
case-13 | pass→pass | 13,633 | 20,544 | +51% | 1 | 1 | 0% | 2,049 | 4,814 | +135% | 0 | 0 | — |
case-14 | fail→pass | 16,140 | 17,429 | +8% | 1 | 1 | 0% | 2,326 | 4,173 | +79% | 0 | 0 | — |
case-15 | fail→pass | 17,349 | 8,927 | -49% | 1 | 1 | 0% | 1,862 | 3,568 | +92% | 0 | 0 | — |
case-16 | fail→pass | 19,403 | 12,434 | -36% | 1 | 1 | 0% | 2,302 | 3,485 | +51% | 0 | 0 | — |
case-17 | fail→fail | 9,190 | 10,030 | +9% | 1 | 1 | 0% | 1,610 | 3,158 | +96% | 0 | 0 | — |
case-18 | fail→pass | 19,369 | 20,803 | +7% | 1 | 1 | 0% | 2,440 | 4,727 | +94% | 0 | 0 | — |
case-19 | fail→pass | 20,386 | 17,524 | -14% | 1 | 1 | 0% | 2,480 | 5,025 | +103% | 0 | 0 | — |
case-20 | fail→pass | 18,967 | 18,453 | -3% | 1 | 1 | 0% | 2,129 | 4,155 | +95% | 0 | 0 | — |
case-21 | fail→pass | 17,680 | 11,577 | -35% | 1 | 1 | 0% | 2,231 | 3,312 | +48% | 0 | 0 | — |
case-22 | fail→pass | 13,917 | 9,712 | -30% | 1 | 1 | 0% | 1,346 | 2,969 | +121% | 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 +50 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.