菜单
Read and extract content from PDF files using multiple methods. LOAD THIS SKILL WHEN: User says "讀取 PDF", "read PDF", "打開論文", "看這篇" | has PDF file path | needs to extract text from PDF. CAPABILITIES: PDF to markdown conversion, text extraction, figure detection, citation extraction.
Validate that written content accurately reflects source materials. LOAD THIS SKILL WHEN: User says "驗證", "validate", "檢查準確性", "核對" | after writing notes/reports | ensuring content accuracy. CAPABILITIES: Source-target comparison, citation verification, fact checking, consistency check.
Filter, validate and organize literature search results. LOAD THIS SKILL WHEN: User has search results and needs to "過濾", "篩選", "filter", "選擇文獻" | reviewing search results | narrowing down papers. CAPABILITIES: Relevance filtering, duplicate detection, inclusion/exclusion criteria, Zotero integration.
Complete literature retrieval capability combining search and filter skills. LOAD THIS SKILL WHEN: User needs "文獻檢索", "找文獻", "retrieve literature", "系統性搜尋" | starting systematic review | comprehensive literature search. CAPABILITIES: Multi-database search, MeSH expansion, quality filtering, PRISMA-compliant workflow. COMPOSITE SKILL: Combines literature-search + literature-filter.
Search and retrieve academic literature from multiple databases. LOAD THIS SKILL WHEN: User says "搜尋文獻", "search literature", "find papers", "PubMed", "學術搜尋" | needs to find research articles | starting literature review. CAPABILITIES: PubMed search, MeSH term expansion, citation metrics, batch retrieval.
Write structured notes from source materials following academic standards. LOAD THIS SKILL WHEN: User says "寫筆記", "write notes", "摘要", "summarize", "整理重點" | has source content to summarize | creating reading notes. CAPABILITIES: Structured summaries, key point extraction, citation formatting, template-based writing.
Complete report writing capability combining reading, note-taking, validation, and formatting. LOAD THIS SKILL WHEN: User needs "寫報告", "write report", "整理成報告", "產出文件" | has source materials to summarize | creating academic document. CAPABILITIES: PDF reading, structured notes, content validation, format checking. COMPOSITE SKILL: Combines pdf-reader + note-writer + content-validator + report-formatter.
| name | pdf-reader |
| description | Read and extract content from PDF files using multiple methods. LOAD THIS SKILL WHEN: User says "讀取 PDF", "read PDF", "打開論文", "看這篇" | has PDF file path | needs to extract text from PDF. CAPABILITIES: PDF to markdown conversion, text extraction, figure detection, citation extraction. |
使用多種方法讀取 PDF 檔案並轉換為結構化文字，支援本地檔案和網路 URL。
| Tool | 用途 | 參數 |
|---|---|---|
convert_to_markdown | 將 PDF 轉換為 Markdown | uri (file:// 或 http://) |
| Tool | 用途 | 參數 |
|---|---|---|
get_article_fulltext_links | 取得全文連結 | pmid |
# scripts/pdf_utils.py
import fitz # PyMuPDF
def extract_text(pdf_path: str) -> str:
"""提取 PDF 文字"""
doc = fitz.open(pdf_path)
text = ""
for page in doc:
text += page.get_text()
return text
def extract_images(pdf_path: str, output_dir: str) -> list:
"""提取 PDF 圖片"""
# ...實作
pass
def extract_references(pdf_path: str) -> list:
"""提取參考文獻"""
# ...實作
pass
用戶提供本地路徑: C:\papers\article.pdf
↓
convert_to_markdown(uri="file:///C:/papers/article.pdf")
↓
返回 Markdown 格式文字
用戶提供 URL 或 PMID
↓
(如果是 PMID) get_article_fulltext_links(pmid)
↓
取得 PDF URL
↓
convert_to_markdown(uri="https://...")
↓
返回 Markdown 格式文字
# 處理多個 PDF
pdf_files = [
"C:/papers/paper1.pdf",
"C:/papers/paper2.pdf",
"C:/papers/paper3.pdf"
]
for pdf in pdf_files:
content = convert_to_markdown(uri=f"file:///{pdf}")
save_to_notes(content, pdf)
# Checkpoint: 記錄處理進度
update_checkpoint(processed=pdf)
## PDF 讀取結果
**檔案**: article.pdf
**頁數**: 12
**處理時間**: 3.2s
---
### 提取內容
# [論文標題]
## Abstract
[摘要內容...]
## Introduction
[介紹內容...]
## Methods
[方法內容...]
## Results
[結果內容...]
## Discussion
[討論內容...]
## References
1. Author et al. (2024). Title...
2. ...
---
### 圖表清單
| 圖表 | 頁碼 | 說明 |
|------|------|------|
| Figure 1 | p.3 | Study flowchart |
| Table 1 | p.5 | Baseline characteristics |
範例 1：本地 PDF
用戶：「讀取 C:\papers\remimazolam-study.pdf」
執行：convert_to_markdown(uri="file:///C:/papers/remimazolam-study.pdf")
範例 2：從 PMID 取得 PDF
用戶：「讀取 PMID 38353755 的全文」
執行：
1. get_article_fulltext_links(pmid="38353755")
2. convert_to_markdown(uri=pmc_url)
範例 3：批次讀取
用戶：「讀取 papers/ 資料夾中的所有 PDF」
執行：
1. 列出資料夾中的 PDF 檔案
2. 迴圈處理每個檔案
3. 建立 Checkpoint 追蹤進度
| 錯誤 | 處理方式 |
|---|---|
| PDF 加密 | 提示用戶提供密碼或使用其他版本 |
| 圖片 PDF (掃描) | 提示需要 OCR，建議使用其他工具 |
| 網路錯誤 | 重試 3 次，失敗後提示用戶下載 |
note-writer - 撰寫筆記report-writing - 組合技能：讀取 + 撰寫