create-pdf-fixture
Create complete PDF test fixtures combining ReportLab tables and AI-generated images. Orchestrates create-table and create-image skills. Creates PDFs designed to test/expose extractor bugs.
ソース情報
- リポジトリ
- grahama1970/agent-stack-public
- ソースの最終更新活動
- 2026年9月24日 15:51
- 検出された SKILL.md の言語
- 英語
- スター
- 0
- フォーク
- 0
インストール方法
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
ソースファイルを確認
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
ファイルエクスプローラー
5 ファイルSKILL.md を表示中
SKILL.md
ソースの指示 · 読み取り専用プレビュー- name
- create-pdf-fixture
- description
- Create complete PDF test fixtures combining ReportLab tables and AI-generated images. Orchestrates create-table and create-image skills. Creates PDFs designed to test/expose extractor bugs.
- allowed-tools
- Bash, Read, Write
- triggers
- ["create pdf fixture","create test pdf","generate pdf fixture","make test pdf","pdf for testing","extractor test fixture","bug reproduction pdf"]
- metadata
- {"short-description":"Create PDF test fixtures (tables + images)"}
- provides
- ["create-pdf-fixture"]
- composes
- ["task-monitor","agentic-evals"]
- disciplines
- ["extraction","evaluation-quality"]
# create-pdf-fixture
Generate complete PDF test fixtures that combine:
- **Proper ReportLab tables** via `create-table` (detectable by Marker/Camelot)
- **AI-generated images** via `create-image` (diagrams, decorative elements)
- **Text content** with various formatting challenges
## Why This Exists
Creating test PDFs that properly exercise extractors requires:
1. Tables built with ReportLab (not raw drawing commands)
2. Images that test VLM classification (decorative vs data)
3. Edge cases: empty sections, malformed titles, nested structures
This skill orchestrates sibling skills for modular, reusable fixtures.
## Quick Start
```bash
cd .pi/skills/create-pdf-fixture
# Generate the extractor bug reproduction fixture
uv run generate.py extractor-bugs --output test.pdf
# Generate a simple fixture
uv run generate.py simple --output simple_test.pdf
# List available presets
uv run generate.py list-presets
# Verify table detection
uv run generate.py verify test.pdf
```
## Presets
| Preset | Description | Tests |
|--------|-------------|-------|
| `extractor-bugs` | Reproduces known extractor issues | Empty sections, false tables, malformed titles |
| `simple` | Basic PDF with table and text | Basic extraction |
## Sibling Skills Used
| Skill | Purpose |
|-------|---------|
| `create-table` | Creates ReportLab tables (Marker-detectable) |
| `create-image` | AI-generated images with caching |
Cached images from `create-image/cached_images/`:
- `decorative.png` - Cover illustration
- `flowchart.png` - Process diagram
- `network_arch.png` - Architecture diagram
## Legacy Usage (Still Supported)
```bash
# Via wrapper (uses extractor project)
./run.sh --example --name test_fixture
./run.sh --spec content_spec.json --name my_fixture
```
## JSON Spec Format (Legacy)
```json
{
"style": "standard",
"sections": [
{
"title": "1. Requirements",
"level": 1,
"content": [
{"type": "text", "text": "This document describes requirements."},
{"type": "table", "columns": ["ID", "Name"], "rows": [["1", "Alice"]]},
{"type": "figure", "description": "Architecture diagram"}
]
}
]
}
```
## Output
- `extractor_bugs_fixture.pdf` - Cached in `cached_fixtures/`
- Custom output via `--output` flag
## Dependencies
```toml
dependencies = [
"pymupdf>=1.23.0",
"reportlab>=4.0.0",
"typer>=0.9.0",
"pillow>=10.0.0",
]
```
GitHubで見る