Document Briefing Skill
Overview
This skill creates structured, comprehensive document briefings from various input types. Core principle: Summary as translation, not compression. Retain sufficient detail—it's better to err on the side of including more information than over-condensing.
When to Use
Use this skill when the user wants to:
- Summarize a URL/webpage into a briefing
- Extract and summarize content from PDF files
- Summarize Word documents (.docx)
- Create briefings from Markdown or text files
- Convert any document into a structured briefing format
Input Types
| Type | Handler |
|---|---|
| URLs | Use WebFetch to get content |
| PDF (.pdf) | Use pdf skill to extract text |
| Word (.docx) | Use docx skill to extract text |
| Markdown (.md) | Read file directly |
| Text (.txt) | Read file directly |
Output Format
The briefing follows this structure:
# [Document Title]
## 📋 基本信息
- 来源类型:
- 获取时间:
- 文档页数/字数:
## 🎯 核心观点
(用 1-3 句话概括文档核心主张)
## 📌 关键发现
1. 发现一
2. 发现二
3. ...
## 📖 详细摘要
### 背景/引言
...
### 主体内容
(按原文结构分段总结,保留足够细节)
### 结论
...
## 💡 分析与洞察
(可选:技能的独立分析视角)
## 📝 待处理事项
- [ ] 待确认的点 1
- [ ] 待确认的点 2
## 🔗 相关资源
(从文档中提取的链接、引用等)
Processing Workflow
Step 1: Identify Input Type
- If URL → fetch with WebFetch
- If PDF → use pdf skill to extract text
- If Word → use docx skill to extract text
- If Markdown/Text → read file directly
Step 2: Extract Content
- Get full text content from the document
- Note metadata (title, page count if available, source URL)
Step 3: Create Briefing
- Title: Extract or infer from content
- 基本信息: Document source type, fetch time, approximate length
- 核心观点: 1-3 sentence summary of main argument
- 关键发现: Numbered list of key findings (3-7 items)
- 详细摘要: Full summary with background, body, conclusion
- 分析与洞察: Optional - your analysis as an expert reader
- 待处理事项: Points that need verification or follow-up
- 相关资源: Extract URLs, citations, references from source
Step 4: Quality Check
Before delivering, verify:
- All sections are present
- Key information is retained (not over-condensed)
- Original structure is reflected in summary
- Links and resources are captured
- Markdown formatting is correct
Quality Guidelines
Do:
- Preserve specific numbers, names, dates, and technical details
- Maintain the original document's logical structure
- Include relevant quotes when they add value
- Note any ambiguities or unclear points for follow-up
- Extract all relevant URLs and references
Don't:
- Over-condense to the point of losing important detail
- Omit negative findings or contradictions
- Add information not in the source
- Use vague summaries like "discusses various topics"
Tips
- For long documents, process each section/chapter recursively but output as ONE unified briefing
- When unsure about detail level, include more rather than less
- Always preserve technical terms, acronyms, and specific data
- Note the document's tone and intended audience if apparent
Document Type Adaptation
Different document types require different emphasis:
| Document Type | Emphasis |
|---|---|
| 论文/研究报告 | 方法论、数据来源、结论、研究价值 |
| 新闻报道 | 5W1H、时间线、关键事实 |
| 商业报告 | 关键数据、建议、行动项、ROI |
| 技术文档 | 架构、API、使用方法、限制 |
| 政策文件 | 适用范围、权利义务、生效时间 |
Handling Long Documents
For documents >20 pages or with clear章节结构:
- Internal recursive processing: 逐章/逐节提取和处理内容,记录每个章节的核心观点和关键发现
- Single unified output: 最终整合成一份完整的简报,不是多份
- 章节映射: 在"详细摘要"部分按原文档章节结构组织内容
- 处理记录 (可选): 可在"📋 基本信息"中添加处理深度说明
处理策略:递归处理全文档 X 章/节,整合为单份简报