菜单
Converts PDF files to structured Markdown with automatic mode selection. Supports simple text extraction (fast) and complex layouts with code/tables (vision). Triggers when the user wants to convert a PDF document, extract PDF content, or transform PDF to Markdown format.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Browser automation with a persistent session, driven by the Playwright Python API. Use when users ask to navigate websites, fill forms, take screenshots, extract web data, test web apps, or automate browser workflows. Trigger phrases include "go to [url]", "click on", "fill out the form", "take a screenshot", "scrape", "automate", "test the website", "log into", or any browser interaction request.
Whole-repo audit for over-engineering. Like ponytail-review, but scans the entire codebase instead of a diff: a ranked list of what to delete, simplify, or replace with stdlib/native equivalents. Use when the user says "audit this codebase", "audit for over-engineering", "what can I delete from this repo", "find bloat", "ponytail-audit", or "/core:ponytail-audit". One-shot report, does not apply fixes.
Harvest every `ponytail:` comment in the codebase into a debt ledger, so the deliberate shortcuts and deferrals ponytail leaves behind get tracked instead of rotting into "later means never". Use when the user says "ponytail debt", "/core:ponytail-debt", "what did ponytail defer", "list the shortcuts", "ponytail ledger", or "what did we mark to do later". One-shot report, changes nothing.
Code review focused exclusively on over-engineering. Finds what to delete: reinvented standard library, unneeded dependencies, speculative abstractions, dead flexibility. One line per finding: location, what to cut, what replaces it. Use when the user says "review for over-engineering", "what can we delete", "is this over-engineered", "simplify review", or invokes /core:ponytail-review. Complements correctness-focused review, this one only hunts complexity.
Forces the laziest solution that actually works, simplest, shortest, most minimal. Channels a senior dev who has seen everything: question whether the task needs to exist at all (YAGNI), reach for the standard library before custom code, native platform features before dependencies, one line before fifty. Supports intensity levels: lite, full (default), ultra. Use whenever the user says "ponytail", "be lazy", "lazy mode", "simplest solution", "minimal solution", "yagni", "do less", or "shortest path", and whenever they complain about over-engineering, bloat, boilerplate, or unnecessary dependencies.
Ingests a new source (web clipping, AI chat export, raw note, transcript) into an Obsidian-based LLM knowledge wiki. Files the source under _raw/, distills it into one or more _wiki/ pages, updates index and log, and pulls bidirectional cross-references. Triggers when the user asks to "ingest this source", "digest this clipping into the wiki", or runs /wiki:ingest while working in a configured wiki vault.
| name | pdf-to-markdown |
| description | Converts PDF files to structured Markdown with automatic mode selection. Supports simple text extraction (fast) and complex layouts with code/tables (vision). Triggers when the user wants to convert a PDF document, extract PDF content, or transform PDF to Markdown format. |
Convert PDF documents to structured Markdown with automatic mode selection.
When the user requests PDF conversion:
Ask the user which mode to use:
| Mode | Best For | Speed |
|---|---|---|
| fast | Simple text documents, reports | Very fast |
| vision | Complex layouts, code, tables, scans | Medium |
Run the converter script:
python ${CLAUDE_PLUGIN_ROOT}/skills/pdf-to-markdown/scripts/pdf_converter.py "<pdf_path>" --mode fast
The script outputs:
<name>.md - Markdown file with extracted text<name>_images/ - Extracted images (if any)Report the result to the user.
Step 1: Convert PDF to page images:
python ${CLAUDE_PLUGIN_ROOT}/skills/pdf-to-markdown/scripts/pdf_converter.py "<pdf_path>" --mode vision [--pages START-END] [--dpi 150]
This creates <name>_pages/ with PNG images.
Step 2: Read each page image using the Read tool and analyze the content.
Step 3: For each page, generate Markdown following these rules:
java, python, etc.)[Image: description]Step 4: Combine all pages into a single Markdown file:
# Document Title
<!-- Page 1 -->
[page 1 content]
---
<!-- Page 2 -->
[page 2 content]
Step 5: Write the final Markdown file using the Write tool.
Step 6: Clean up temporary page images (ask user first).
User: Convert /Users/me/report.pdf to Markdown
Claude: I will convert the PDF. Which mode do you prefer?
- **fast**: Quick text extraction (for simple documents)
- **vision**: Image analysis (for complex layouts, code, tables)
User: fast
Claude: [Runs script, reports result]
Done! Created:
- /Users/me/report.md
- /Users/me/report_images/ (3 images)
User: Convert java-book.pdf to Markdown, contains a lot of code
Claude: Since the document contains code, I recommend vision mode.
Should I convert the entire PDF or only specific pages?
User: Pages 10-30
Claude: [Converts pages to images]
[Reads each page image]
[Generates structured Markdown with code blocks]
[Writes output file]
Done! Created: java-book.md (21 pages)
Should I delete the temporary page images?
User: /pdf-to-markdown /path/to/document.pdf
Claude: [Asks for mode preference]
[Executes conversion]
[Reports result]
Ensure dependencies are installed before first use:
# Fast mode
pip install PyMuPDF Pillow --break-system-packages
# Vision mode (additional)
pip install pdf2image --break-system-packages
brew install poppler # macOS
| Option | Description | Default |
|---|---|---|
--mode fast | PyMuPDF text extraction | Default |
--mode vision | Prepare images for analysis | - |
--pages START-END | Process specific pages | All |
--dpi N | Image resolution | 150 |
--no-images | Skip image extraction (fast) | - |
--use-api | Direct Anthropic API (vision mode) | Off |
--model NAME | Claude model for --use-api calls | claude-opus-4-7 |
Note on
--model: The default isclaude-opus-4-7for high-resolution image handling. Set--model claude-sonnet-4-6for lower-cost runs on simpler documents. Direct API usage (--use-api) bills the user'sANTHROPIC_API_KEYper page.