阅读学术论文,评估主要论点、证据、研究方法、创新性及其与相关研究的关系,具体指出优点与不足,并整理成包含改进建议的评审。
- 了解论文成果与局限
- 起草同行评审意见
- 检查实验设计与可复现性
围绕指定主题,从多个网页来源调查新闻与资料,整理成附有摘要、解读和来源链接的新闻简报,并根据读者和发布频率调整结构与文风,输出Markdown。
Use this skill when the user requests to generate, create, write, or draft a newsletter, email digest, weekly roundup, industry briefing, or curated content summary. Supports topic-based research, content curation from multiple sources, and professional formatting for email or web distribution. Trigger on requests like "create a newsletter about X", "write a weekly digest", "generate a tech roundup", or "curate news about Y".
围绕指定主题调查多个网页来源,将内容摘要与独立解读结合,编写新闻简报。整理主要新闻、趋势分析、简短消息、工具和参考资料,形成附有来源链接的Markdown文稿。支持每日摘要、每周汇总、专题深入解读及行业报告,并根据目标读者调整文风和说明的详细程度。
适合定期向读者传递某个领域的动态,或将多篇文章整理为适合读者阅读的内容。可按目的选择结构,既能面向普通读者提供易懂的汇总,也能为管理层撰写着重分析影响的解读。持续发布时,会保持简报名称和标题结构的一致性。
调查使用网页搜索和获取文章正文的web_fetch。输出为Markdown,转换成HTML需另用转换工具。遵循避免概括付费文章等读者无法访问内容的方针。
本介绍由 AI(OpenAI Codex)根据公开的 SKILL.md 撰写。准确的规格请查看下方原文。
AI 仅根据 SKILL.md 的内容,从 4 个方面进行了评估。星标数和热度不计入评估。
总分按每项 A=25 分、B=15 分、C=5 分相加。 由 AI(OpenAI Codex)于 2026年10月21日 评估。仅供参考,安装前请同时查看原文。
安装前可以查看交给 AI 代理的指令内容。
This skill generates professional, well-researched newsletters that combine curated content from multiple sources with original analysis and commentary. It follows modern newsletter best practices from publications like Morning Brew, The Hustle, TLDR, and Benedict Evans to produce content that is informative, engaging, and actionable.
The output is a complete, ready-to-publish newsletter in Markdown format, suitable for email distribution platforms, web publishing, or conversion to HTML.
Always load this skill when:
Identify the key parameters:
| Parameter | Description | Default |
|---|---|---|
| Topic(s) | Primary subject area(s) to cover | Required |
| Format | Daily digest, weekly roundup, deep-dive, or industry briefing | Weekly roundup |
| Target Audience | Technical, executive, general, or niche community | General |
| Tone | Professional, conversational, witty, or analytical | Conversational-professional |
| Length | Short (5-min read), medium (10-min), long (15-min+) | Medium |
| Sections | Number and type of content sections | 4-6 sections |
| Frequency Context | One-time or part of a recurring series | One-time |
Based on the format, select the appropriate structure:
Daily Digest Structure:
1. Top Story (1 item, detailed)
2. Quick Hits (3-5 items, brief)
3. One Stat / Quote of the Day
4. What to Watch
Weekly Roundup Structure:
1. Editor's Note / Intro
2. Top Stories (2-3 items, detailed)
3. Trends & Analysis (1-2 items, original commentary)
4. Quick Bites (4-6 items, brief summaries)
5. Tools & Resources (2-3 items)
6. One More Thing / Closing
Deep-Dive Structure:
1. Introduction & Context
2. Background / Why It Matters
3. Key Developments (detailed analysis)
4. Expert Perspectives
5. What's Next / Implications
6. Further Reading
Industry Briefing Structure:
1. Executive Summary
2. Market Developments
3. Company News & Moves
4. Product & Technology Updates
5. Regulatory & Policy Changes
6. Data & Metrics
7. Outlook
Conduct thorough research using web search. The quality of the newsletter depends directly on the quality and recency of research.
Search Strategy:
# Current news and developments
"[topic] news [current month] [current year]"
"[topic] latest developments"
"[topic] announcement this week"
# Trends and analysis
"[topic] trends [current year]"
"[topic] analysis expert opinion"
"[topic] industry report"
# Data and statistics
"[topic] statistics [current year]"
"[topic] market data latest"
"[topic] growth metrics"
# Tools and resources
"[topic] new tools [current year]"
"[topic] open source release"
"best [topic] resources [current year]"
IMPORTANT: Always check
<current_date>to ensure search queries use the correct temporal context. Never use hardcoded years.
Evaluate each source and curate the best content:
| Criterion | Priority |
|---|---|
| Recency | Prefer content from the last 7-30 days |
| Authority | Prioritize primary sources, official announcements, established publications |
| Uniqueness | Select stories that offer fresh perspective or are underreported |
| Relevance | Every item must clearly connect to the newsletter's stated topic(s) |
| Actionability | Prefer content readers can act on (tools, insights, strategies) |
| Diversity | Mix of news, analysis, data, and practical resources |
For key stories, use web_fetch to read full articles and extract:
Every newsletter starts with a consistent header:
# [Newsletter Name]
*[Tagline or description] — [Date]*
---
[Optional: One-sentence preview of what's inside]
Top Stories / Featured Items:
Quick Bites / Brief Items:
Analysis / Commentary Sections:
| Principle | Implementation |
|---|---|
| Scannable | Use headers, bold text, bullet points, and short paragraphs |
| Engaging | Lead with the most interesting angle, not chronological order |
| Concise | Every sentence earns its place — cut filler ruthlessly |
| Accurate | Every fact is sourced, every number is verified |
| Attributive | Always credit original sources with inline links |
| Human | Write like a knowledgeable friend, not a press release |
Tone Calibration by Audience:
| Audience | Tone | Example |
|---|---|---|
| Technical | Precise, no jargon explanations, assumed expertise | "The new API supports gRPC streaming with backpressure handling via flow control windows." |
| Executive | Impact-focused, bottom-line, strategic | "This acquisition gives Company X a 40% market share in the enterprise segment, directly threatening Incumbent Y's pricing power." |
| General | Accessible, analogies, explains concepts | "Think of it like a universal translator for data — it lets any app talk to any database without learning a new language." |
Combine all sections into the final document following the chosen structure template.
Every newsletter ends with:
---
*[Newsletter Name] is [description of what it is].*
*[How to subscribe/share/give feedback]*
*Sources: All links are provided inline. This newsletter curates and summarizes
publicly available information with original commentary.*
Before finalizing, verify:
# [Newsletter Name]
*[Tagline] — [Full date, e.g., April 4, 2026]*
---
[Preview sentence: "This week: [topic 1], [topic 2], and [topic 3]."]
## 🔥 Top Stories
### [Headline 1]
[Hook — why this matters in 1-2 sentences.]
[Body — 2-4 paragraphs covering key facts, context, and implications.]
**Why it matters:** [1 paragraph connecting to reader's interests or industry impact.]
📎 [Source: Publication Name](URL)
### [Headline 2]
[Same structure as above]
## 📊 Trends & Analysis
### [Trend Title]
[Original commentary on an emerging trend, backed by data from research.]
[Key data points presented clearly — consider inline stats or a brief comparison.]
**The bottom line:** [One-sentence takeaway.]
## ⚡ Quick Bites
- **[Headline]** — [2-3 sentence summary with key takeaway.] [Source](URL)
- **[Headline]** — [2-3 sentence summary.] [Source](URL)
- **[Headline]** — [2-3 sentence summary.] [Source](URL)
- **[Headline]** — [2-3 sentence summary.] [Source](URL)
## 🛠️ Tools & Resources
- **[Tool/Resource Name]** — [What it does and why it's useful.] [Link](URL)
- **[Tool/Resource Name]** — [Description.] [Link](URL)
## 💬 One More Thing
[Closing thought, insightful quote, or forward-looking statement.]
---
*[Newsletter Name] curates the most important [topic] news and analysis.*
*Found this useful? Share it with a colleague.*
*All sources are linked inline. Views and commentary are original.*
After generation:
/mnt/user-data/outputs/newsletter-{topic}-{date}.mdpresent_files tooldeep-research skill for comprehensive topic coverage — load both for newsletters requiring deep analysis<current_date> for temporal context in searches and date references in the newsletter还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
阅读学术论文,评估主要论点、证据、研究方法、创新性及其与相关研究的关系,具体指出优点与不足,并整理成包含改进建议的评审。
检测可能阻塞Python异步处理的输入输出调用,并推动修复和补充测试。验证问题重现时测试确实失败、恢复修复后通过,以监测问题再次出现。
通过对话梳理工作、偏好及希望交由AI判断的范围,创建或更新规定AI伙伴行为的SOUL.md,并明确说话方式和表达异议的规则。
根据数据特点,从26种图表和示意图中选择合适的形式并生成图片,可展示时间变化、比较、分布和关联,并返回图片链接及生成时使用的完整配置。
通过HTTP API向运行中的DeerFlow发送调研和分析请求,获取回答、继续对话并上传资料。还可查看模型、技能、智能体列表及服务运行状态。
阅读实际源代码,按项目规模编写或完善包含安装步骤与用法示例的README、API文档、架构图、开发者指南及变更日志。