# PDF

> 使用本地开源工具创建、检查、提取、合并、拆分、旋转、渲染与校验 PDF 文件。当需求涉及 PDF 输入或输出、页面级转换、文本或表格提取，或需要对生成的 PDF 做视觉核验时使用。

- Skill: `wecode-ai/pdf` (Agent Skill, multi-file: 7 files)
- Install (CLI): `npx skillmds@latest add wecode-ai/pdf`
- Raw SKILL.md: https://api.skillmd.com/api/skills/wecode-ai/pdf/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: wecode-ai (https://skillmd.com/u/wecode-ai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/wecode-ai/pdf

---


# PDF

Use the deterministic PDF CLI for common creation and transformation tasks. Preserve source files and write transformed PDFs to new paths.

## Runtime

Prefer workspace dependency Python. Otherwise use Python 3:

On macOS/Linux use `python3`; on Windows PowerShell use `py -3` (or the
workspace dependency loader's absolute `python.exe` path). On Windows set the
skill directory with `$env:SKILL_DIR = '<skill-directory>'`. Every `python3`
example below uses this platform-specific launcher substitution.

```bash
python3 "$SKILL_DIR/scripts/bootstrap.py" <command> [arguments]
```

```powershell
py -3 "$env:SKILL_DIR\scripts\bootstrap.py" <command> [arguments]
```

Missing packages are installed from the hashed lock into `~/.wegent-executor/plugin-envs/wework-public/pdf/`, never globally.

## Workflow

1. Run `inspect --text` or `extract` to understand an input PDF.
2. For new PDFs, author a JSON specification following [spec.md](references/spec.md), then run `create`.
3. Use `merge`, `split`, or `rotate` for page-level transformations.
4. Run `validate` on the final PDF.
5. Run `render` and inspect every output page image before delivery when layout matters.

## Commands

```bash
python3 "$SKILL_DIR/scripts/bootstrap.py" inspect --input source.pdf --text --output inspection.json
python3 "$SKILL_DIR/scripts/bootstrap.py" extract --input source.pdf --tables --output extraction.json
python3 "$SKILL_DIR/scripts/bootstrap.py" create --spec report.json --output report.pdf
python3 "$SKILL_DIR/scripts/bootstrap.py" merge --inputs first.pdf second.pdf --output merged.pdf
python3 "$SKILL_DIR/scripts/bootstrap.py" split --input merged.pdf --pages '1-3,5' --output selection.pdf
python3 "$SKILL_DIR/scripts/bootstrap.py" rotate --input source.pdf --degrees 90 --pages '2' --output rotated.pdf
python3 "$SKILL_DIR/scripts/bootstrap.py" validate --input result.pdf
python3 "$SKILL_DIR/scripts/bootstrap.py" render --input result.pdf --output-dir rendered
```

## Quality rules

- Do not overwrite input PDFs.
- Preserve page order intentionally and use 1-based page selections.
- Treat extracted text as an interpretation of PDF drawing instructions; verify tables and reading order against rendered pages.
- Warn when a PDF is encrypted or content extraction is incomplete.
- A structurally valid PDF still requires visual inspection when it is user-facing.