Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Xay Brand Voice Document — nen tang cua content cluster: brand personality 3 tinh tu, tone of voice 5 chieu, vocabulary (dung/tranh/banned words), cau truc cau, 5 vi du before/after, platform adaptation va checklist duyet copy. Kich hoat khi nhac 'brand voice', 'tone of voice', 'giong brand', 'cach viet cua brand', 'brand personality'.
.claude/skills/minhnv0807-35-brand-voice/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 60% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 75% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 57% | 0% |
> Skill nen tang cua content cluster. Moi skill content khac (36-content-brief, 37-caption-social, 04-script-video, 05-copy-quang-cao, 14-email-marketing, 38-seeding-plan) deu doc Brand Voice Document truoc khi viet. Khong co brand voice thi moi output deu thieu nhat quan, du ky thuat dung.
Doc .agents/product-marketing-context.md va brand guideline neu co. Neu thieu, hoi toi da 4 cau:
09-insight-khach-hang de co customer language truoc khi chot tu vung.Chon dung 3 tinh tu cot loi, moi tinh tu kem 1 cau giai thich tai sao no phan anh dung brand. Khong chon qua 3 — nhieu hon la loang.
Vi du minh hoa (nganh tu van marketing):
Kem theo 3 dieu brand tuyet doi KHONG muon bi nhin nhan (vi du: khong "sang chanh xa cach", khong "ban hang bat chap", khong "ly thuyet suong").
Voi moi chieu, viet ca 2 cot "Brand la..." va "Brand khong la..." — cot thu 2 quan trong khong kem cot dau vi no chan brand drift.
| Chieu | Brand la... | Brand khong la... | |-------|-------------|-------------------| | Nang luong | soi noi / diem tinh / am ap] | ...] | | Do chinh thuc | than mat / ban-chuyen-nghiep / trang trong] | ...] | | Cach xung ho | minh-ban / chung toi-ban / toi-ban] | ...] | | Cach dung so lieu / bang chung | luon kem so cu the / ke chuyen truoc, so sau] | ...] | | Cach xu ly objection | thua nhan truoc, giai thich sau / phan bien bang bang chung] | ...] |
Tra loi du 5 cau hoi:
Moi vi du gom: ban chua dung tone, ban dung tone, va 1 dong giai thich diem khac biet. Bat buoc du 5 tinh huong:
Day la phan writer dung nhieu nhat — viet that theo brand, khong de placeholder.
Giong goc giu nguyen — chi dieu chinh format va do dai:
| Kenh | Dieu chinh so voi goc | Giu nguyen | |------|----------------------|------------| | Facebook Page | Dai hon, co the dung emoji | Tone, vocabulary | | TikTok caption | Ngan, co the dung trending phrase | Core message | | Instagram | Visual-first, caption ngan | Brand voice | | Email | Formal hon 1 chut, co greeting | Thang than, actionable | | Zalo OA | Ngan gon, di thang vao loi ich | Tone, cach xung ho |
Ten file: brand-voice-[ten-brand]-[YYYYMMDD].md
markdown# Brand Voice Document — [Ten brand] ## I. Brand Personality - [Tinh tu 1]: [1 cau giai thich] - [Tinh tu 2]: [1 cau giai thich] - [Tinh tu 3]: [1 cau giai thich] - Khong bao gio la: [3 dieu brand khong muon bi nhin nhan] ## II. Tone of Voice | Chieu | Brand la... | Brand khong la... | |-------|-------------|-------------------| | Nang luong | | | | Do chinh thuc | | | | Cach xung ho | | | | Cach dung so lieu / bang chung | | | | Cach xu ly objection | | | ## III. Vocabulary **Dung thuong xuyen:** [tu / cum tu] **Tranh:** [tu / cum tu] **Banned words:** [danh sach cam tuyet doi] **Ten goi chuan noi bo:** [ten san pham / dich vu thong nhat] ## IV. Cau truc cau - Do dai cau trung binh: [ngan / vua / dai] — [ly do] - Cau hoi tu tu: [khi nao dung / khong dung] - Bullet vs paragraph: [ti le + ngu canh] - Pattern mo bai: [Pain / Curiosity / Data / Story] - Pattern ket bai: [CTA / Question / Summary] ## V. Before / After — 5 vi du ### 1. Caption gioi thieu san pham - Before: "[ban chua dung tone]" - After: "[ban dung tone]" - Khac biet: [1 dong] ### 2. Phan hoi comment tieu cuc [cung format] ### 3. Thong bao khuyen mai / uu dai [cung format] ### 4. Chia se kien thuc / edu content [cung format] ### 5. Testimonial / case study [cung format] ## VI. Platform Adaptation | Kenh | Dieu chinh so voi goc | Giu nguyen | |------|----------------------|------------| | | | | ## VII. Checklist duyet copy - [ ] Doc to len — nghe tu nhien khong, hay nhu doc bao cao? - [ ] Co tu nao trong banned list khong? - [ ] Cau dau tien co khien nguoi ta muon doc tiep khong? - [ ] Xoa duoc tu nao ma khong mat nghia khong? Xoa di. - [ ] Khach doc xong biet lam gi tiep khong (CTA ro)? - [ ] Che ten brand di — doc van nhan ra la minh khong?
09-insight-khach-hang: chay truoc — customer language tu review/comment that la nguyen lieu cho Vocabulary.36-content-brief: moi brief tham chieu brand voice trong phan tone.37-caption-social, 04-script-video, 05-copy-quang-cao, 14-email-marketing: doc tai lieu nay truoc khi viet output.brand-voice-[ten-brand]-[YYYYMMDD].md| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 46,722 | 42,549 | -9% | 1 | 1 | 0% | 6,712 | 9,321 | +39% | 0 | 0 | — |
case-02 | fail→pass | 54,698 | 35,977 | -34% | 1 | 1 | 0% | 8,263 | 7,522 | -9% | 0 | 0 | — |
case-03 | fail→fail | 48,211 | 41,830 | -13% | 1 | 1 | 0% | 6,759 | 8,182 | +21% | 0 | 0 | — |
case-04 | fail→fail | 39,045 | 40,416 | +4% | 1 | 1 | 0% | 6,554 | 8,388 | +28% | 0 | 0 | — |
case-05 | fail→fail | 41,451 | 31,940 | -23% | 1 | 1 | 0% | 6,014 | 7,803 | +30% | 0 | 0 | — |
case-06 | fail→pass | 30,501 | 32,304 | +6% | 1 | 1 | 0% | 5,131 | 8,195 | +60% | 0 | 0 | — |
case-07 | fail→fail | 32,813 | 36,065 | +10% | 1 | 1 | 0% | 5,248 | 8,062 | +54% | 0 | 0 | — |
case-08 | fail→pass | 30,506 | 29,338 | -4% | 1 | 1 | 0% | 3,952 | 6,920 | +75% | 0 | 0 | — |
case-09 | fail→fail | 44,473 | 65,632 | +48% | 1 | 1 | 0% | 6,088 | 8,953 | +47% | 0 | 0 | — |
case-10 | fail→fail | 32,458 | 33,970 | +5% | 1 | 1 | 0% | 5,210 | 8,331 | +60% | 0 | 0 | — |
case-11 | fail→pass | 30,837 | 30,596 | -1% | 1 | 1 | 0% | 4,237 | 6,654 | +57% | 0 | 0 | — |
case-12 | pass→pass | 32,372 | 38,469 | +19% | 1 | 1 | 0% | 5,353 | 7,875 | +47% | 0 | 0 | — |
case-13 | fail→pass | 21,954 | 31,298 | +43% | 1 | 1 | 0% | 3,570 | 6,683 | +87% | 0 | 0 | — |
case-14 | fail→pass | 47,142 | 41,793 | -11% | 1 | 1 | 0% | 6,980 | 8,080 | +16% | 0 | 0 | — |
case-15 | fail→pass | 28,170 | 42,449 | +51% | 1 | 1 | 0% | 3,644 | 7,277 | +100% | 0 | 0 | — |
case-16 | fail→pass | 52,321 | 48,061 | -8% | 1 | 1 | 0% | 7,879 | 7,418 | -6% | 0 | 0 | — |
case-17 | fail→pass | 39,740 | 63,701 | +60% | 1 | 1 | 0% | 5,638 | 8,139 | +44% | 0 | 0 | — |
case-18 | fail→fail | 38,921 | 41,400 | +6% | 1 | 1 | 0% | 5,068 | 7,539 | +49% | 0 | 0 | — |
case-19 | fail→pass | 45,329 | 41,156 | -9% | 1 | 1 | 0% | 4,517 | 8,545 | +89% | 0 | 0 | — |
case-20 | pass→pass | 27,007 | 26,098 | -3% | 1 | 1 | 0% | 2,143 | 5,610 | +162% | 0 | 0 | — |
case-21 | pass→fail | 73,272 | 62,980 | -14% | 1 | 1 | 0% | 5,124 | 10,649 | +108% | 0 | 0 | — |
case-22 | pass→pass | 19,351 | 42,558 | +120% | 1 | 1 | 0% | 2,354 | 6,595 | +180% | 0 | 0 | — |
case-23 | fail→fail | 11,116 | 30,062 | +170% | 1 | 1 | 0% | 577 | 3,631 | +529% | 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. 23 cases were attempted. The headline lift of +43 percentage points is the difference between those two pass rates over the 23 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.