Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Creative writing skill for creating style skills that teach Claude to write in specific styles. Use when you want to create style guides that Claude can follow when writing long-form content. Creates AI-directive style guides based on analyzing existing writing samples. 트리거: "내 글체로 스킬 만들어줘", "문체 스킬 생성", "글쓰기 스타일 학습", "/cw-style-skill-creator". 산출물이 스킬 파일이 아니라 보이스 프로필 문서면 voice-dna-creator 사용.
.claude/skills/bam-bam-2-cw-style-skill-creator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 25% | 0% |
Create style skills that teach Claude your writing style.
This creates AI instructions (for Claude to read), NOT human documentation (for authors to read).
| AI Instructions | Human Documentation | |-----------------|---------------------| | "When writing X, do Y" | "The story uses X because Y" | | Directive commands | Explanatory descriptions | | Pattern + examples | Analysis + reasoning |
Always ask first:
다음 중 어떤 방식으로 스타일 스킬을 만들까요?
1. 기존 글 샘플 제공
- 1~5장 같은 실제 쓰신 글을 분석해서 패턴 추출
2. 스타일 직접 설명
- "나는 이렇게 쓴다"를 말로 설명
어느 쪽으로 하시겠어요?From existing prose (preferred):
From description:
Use imperative/command form:
✅ "Use short sentences during emotional moments" ✅ "Avoid transition words like '또한', '그리고'" ✅ "Show vulnerability through specific episodes, not declarations" ❌ "The author tends to use short sentences" (analysis, not instruction)
Always include examples:
markdown**Personal anecdotes:** - Always anchor insight to a specific moment/episode - Example: "일요일 저녁이었습니다. 정산을 하니 새벽이었습니다." not "모임이 힘들었습니다."
Pattern + Example format:
markdown**[Pattern name]:** - [Instruction about the pattern] - Example: [Concrete example from the author's writing] - Avoid: [What NOT to do]
Create a SKILL.md file with these sections:
markdown--- name: [author]-writing-style description: Style guide for writing in [author]'s voice. Load this when writing content for [author]. --- # [Author] Writing Style ## Core Voice [2-3 sentences capturing the essential feel] ## Sentence & Paragraph Structure [Directives about length, rhythm, structure] ## Emotional Tone [Directives about vulnerability, directness, energy level] ## Opening Patterns [How chapters/sections typically start] ## Closing Patterns [How sections end - with what kind of beat] ## Vocabulary & Phrases [Signature expressions, words to use, words to avoid] ## Anecdote Style [How personal stories are structured and used] ## What to Avoid [Explicit list of things that feel wrong for this voice]
Master Prose: Overall writing voice, sentence structure, tone Personal Essay: Anecdote usage, vulnerability, insight delivery Social Content: Platform-specific patterns (Threads, LinkedIn) Chapter Structure: Long-form flow, section openings/closings
The ideal workflow for ebooks:
cw-style-skill-creator → analyze 1~5장bambam-writing-style skillbambam-writing-style + humanize-writing| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 14,767 | 4,372 | -70% | 1 | 1 | 0% | 2,480 | 1,315 | -47% | 0 | 0 | — |
case-02 | fail→pass | 13,184 | 3,703 | -72% | 1 | 1 | 0% | 1,910 | 1,202 | -37% | 0 | 0 | — |
case-03 | pass→pass | 10,794 | 3,669 | -66% | 1 | 1 | 0% | 1,560 | 1,178 | -24% | 0 | 0 | — |
case-04 | fail→pass | 11,859 | 4,951 | -58% | 1 | 1 | 0% | 1,588 | 1,352 | -15% | 0 | 0 | — |
case-05 | pass→pass | 11,321 | 6,667 | -41% | 1 | 1 | 0% | 1,582 | 1,733 | +10% | 0 | 0 | — |
case-06 | pass→pass | 13,665 | 9,110 | -33% | 1 | 1 | 0% | 2,069 | 1,920 | -7% | 0 | 0 | — |
case-07 | pass→pass | 13,626 | 13,804 | +1% | 1 | 1 | 0% | 1,827 | 1,966 | +8% | 0 | 0 | — |
case-08 | fail→pass | 11,934 | 6,774 | -43% | 1 | 1 | 0% | 1,580 | 1,208 | -24% | 0 | 0 | — |
case-09 | fail→fail | 14,670 | 12,651 | -14% | 1 | 1 | 0% | 2,247 | 2,727 | +21% | 0 | 0 | — |
case-10 | pass→pass | 17,534 | 10,561 | -40% | 1 | 1 | 0% | 2,409 | 2,074 | -14% | 0 | 0 | — |
case-11 | pass→pass | 14,845 | 9,608 | -35% | 1 | 1 | 0% | 1,966 | 2,301 | +17% | 0 | 0 | — |
case-12 | pass→pass | 17,113 | 13,438 | -21% | 1 | 1 | 0% | 2,686 | 2,511 | -7% | 0 | 0 | — |
case-13 | pass→fail | 6,498 | 4,558 | -30% | 1 | 1 | 0% | 930 | 1,522 | +64% | 0 | 0 | — |
case-14 | pass→pass | 25,376 | 21,108 | -17% | 1 | 1 | 0% | 3,219 | 3,648 | +13% | 0 | 0 | — |
case-15 | pass→pass | 8,623 | 8,520 | -1% | 1 | 1 | 0% | 1,058 | 2,182 | +106% | 0 | 0 | — |
case-16 | pass→pass | 15,818 | 7,478 | -53% | 1 | 1 | 0% | 2,203 | 1,798 | -18% | 0 | 0 | — |
case-17 | pass→pass | 13,344 | 8,575 | -36% | 1 | 1 | 0% | 1,898 | 1,855 | -2% | 0 | 0 | — |
case-18 | pass→pass | 9,031 | 4,773 | -47% | 1 | 1 | 0% | 1,297 | 1,400 | +8% | 0 | 0 | — |
case-19 | fail→fail | 11,855 | 7,911 | -33% | 1 | 1 | 0% | 1,806 | 1,831 | +1% | 0 | 0 | — |
case-20 | fail→pass | 14,508 | 14,858 | +2% | 1 | 1 | 0% | 2,059 | 2,581 | +25% | 0 | 0 | — |
case-21 | fail→pass | 12,400 | 7,477 | -40% | 1 | 1 | 0% | 1,824 | 1,781 | -2% | 0 | 0 | — |
case-22 | pass→pass | 15,475 | 11,189 | -28% | 1 | 1 | 0% | 2,212 | 2,334 | +6% | 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 +23 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.