Code Skill

Code Skill Template

스크립트를 포함하는 스킬용

디렉토리 구조

my-code-skill/
├── SKILL.md
├── reference.md
└── scripts/
    ├── analyze.py
    ├── process.py
    └── validate.py

SKILL.md 예시

---
name: processing-pdfs
description: "Extracts text from PDFs, fills forms, merges documents. Use when working with PDF files or document extraction."
allowed-tools:
  - Bash
  - Read
  - Write
---

# PDF Processing

## Quick start

Extract text:
```bash
python scripts/extract_text.py input.pdf > output.txt

Workflow

Copy this checklist:

Progress:
- [ ] Step 1: Analyze PDF structure
- [ ] Step 2: Extract content
- [ ] Step 3: Validate output

Step 1: Analyze PDF

python scripts/analyze.py input.pdf

Output shows page count, form fields, and structure.

Step 2: Extract content

python scripts/extract_text.py input.pdf > output.txt

Step 3: Validate

python scripts/validate.py output.txt

Fix any issues before proceeding.

Scripts reference

For detailed API: See reference.md

Dependencies

pip install pypdf pdfplumber

---

## scripts/analyze.py 예시

```python
#!/usr/bin/env python3
"""
analyze.py - PDF 구조 분석

Usage:
    python analyze.py input.pdf

Output:
    - Page count
    - Form fields
    - Text/image ratio
"""

import sys
import json
from pypdf import PdfReader

def main():
    if len(sys.argv) != 2:
        print("Usage: python analyze.py input.pdf", file=sys.stderr)
        sys.exit(1)

    pdf_path = sys.argv[1]

    try:
        reader = PdfReader(pdf_path)
    except Exception as e:
        print(f"Error reading PDF: {e}", file=sys.stderr)
        sys.exit(1)

    result = {
        "page_count": len(reader.pages),
        "form_fields": [],
        "has_text": False
    }

    # Form fields
    if reader.get_form_text_fields():
        result["form_fields"] = list(reader.get_form_text_fields().keys())

    # Check for text
    for page in reader.pages:
        if page.extract_text().strip():
            result["has_text"] = True
            break

    print(json.dumps(result, indent=2))

if __name__ == "__main__":
    main()

핵심 포인트

1. 워크플로우 체크리스트: 진행 추적
2. 피드백 루프: 검증 → 수정 → 재검증
3. 스크립트 문서화: Usage, Output 명시
4. 에러 직접 처리: Claude에 떠넘기지 않기
5. 의존성 명시: pip install 명령 포함