hs3180/daily-news-inspiration

Daily News Inspiration

Browse the latest social and tech news daily, extract inspiration, and generate a natural question to simulate real-user interaction.

When to Use This Skill

Use this skill for:

• Generating daily questions based on current news
• Testing disclaude's ability to respond to current events
• Discovering interesting discussion topics from news

Keywords that trigger this skill: "新闻灵感", "每日提问", "时事提问", "新闻提问", "news inspiration", "daily news question"

Core Principle

Natural divergence, not templated questions.

The question should feel like a real user saw the news and naturally wanted to discuss it. Do NOT use preset question templates.

Context Variables

When invoked, you receive:

• Chat ID: Feishu chat ID (from "Chat ID: xxx")
• Message ID: Message ID (from "Message ID: xxx")
• Sender Open ID: Sender's open ID (from "Sender Open ID: xxx")

---

Execution Process

Step 1: Browse Latest News

Use WebSearch to browse today's latest social and tech news. Focus on two categories:

1. Social news (社会新闻): Major events, policy changes, cultural trends
2. Tech news (科技新闻): Product launches, AI breakthroughs, industry trends

Search queries:

WebSearch("今日社会新闻热点 2026")
WebSearch("今日科技新闻 AI 2026")

Step 2: Select One News Item

From the search results, pick one news item that:

• Is genuinely interesting or thought-provoking
• Has potential for deeper discussion
• Is relevant to the user's context (tech-savvy, uses AI tools)
• Is NOT purely negative or sensational

Selection criteria:

• Prefers tech/AI-related news (higher relevance)
• Prefers news with nuance (not just "X happened")
• Avoids purely political or controversial topics
• Avoids news that requires specialized domain knowledge

Step 3: Generate Natural Question

Write a question in a real user's voice, as if they just saw this news and wanted to discuss it.

Question characteristics:

• Natural, conversational tone (not formal or journalistic)
• Shows genuine curiosity or interest
• Open-ended (invites discussion, not yes/no)
• Brief background context (1-2 sentences) so the question is self-contained

Question format:

{1-2 sentences of news background, casual tone}

{Natural follow-up question or curiosity}

Step 4: Send Question

Use the mcp__channel-mcp__send_text tool to send the question to the target chat:

mcp__channel-mcp__send_text({
  text: "{generated question}",
  chatId: "{chatId}"
})

Step 5: Observe and Evaluate (Internal)

After sending, mentally note (do NOT send to user):

• Did disclaude understand the news context?
• Was the response relevant and helpful?
• Any gaps or issues worth reporting?

If significant issues are found, create a GitHub issue for improvement.

---

Question Type Variety

Rotate between different question types across days:

| Type | Description | Example | |------|-------------|---------| | Analysis | Asking for deeper analysis of a trend | "What's the real impact of X?" | | Opinion | Asking for perspective on a development | "What do you think about X's approach?" | | Factual | Asking about specific details | "How does X actually work?" | | Creative | Asking for creative speculation | "What if X was applied to Y?" | | Comparison | Asking to compare approaches | "X vs Y - which makes more sense?" |

---

Quality Guidelines

Good Questions:

• Based on real, current news
• Feels like a real person asking
• Open-ended with room for discussion
• Self-contained (includes enough context)
• Varied in type and topic

Avoid:

• Yes/no questions
• Questions that need specialized knowledge to answer
• Questions about purely negative/tragic events
• Template-like or repetitive questions
• Questions that feel like a test rather than genuine curiosity
• Multiple questions per day (quality over quantity)

---

Example

Search Results:

• "GPT-5 released with multi-modal reasoning"
• "China's new data privacy regulation"
• "Quantum computing breakthrough at Google"

Selected News:

GPT-5 with multi-modal reasoning (tech-relevant, interesting)

Generated Question:

刚看到 GPT-5 发布了，据说支持多模态推理，能同时理解图片和文字的复杂关系了。你觉得这种能力对日常开发工作会有实际影响吗？还是说更多是概念上的进步？

---

安装步骤

1. 收集参数

| Parameter | Required | Default | Description | |-----------|----------|---------|-------------| | {controlChannelChatId} | Yes | — | 目标群组 chatId（当前对话的 chatId） | | {cron} | No | 0 9 * * * | 执行频率（默认每天上午 9:00） |

2. 实例化 Schedule

读取 skill 目录中的 schedule.template.md 模板，替换占位符后写入 workspace：

# 1. 读取模板内容（使用 Read 工具）
模板路径: skills/daily-news-inspiration/schedule.template.md

# 2. 替换所有占位符

| 占位符 | 替换为 | |--------|--------| | {controlChannelChatId} | 实际的目标群组 chatId | | {cron} | 实际的 cron 表达式（默认 0 9 * * *） |

# 3. 使用 Write 工具写入目标文件
目标路径: schedules/daily-news-inspiration/SCHEDULE.md

3. 验证

读取生成的 schedules/daily-news-inspiration/SCHEDULE.md，确认：

• frontmatter 中无未替换的占位符
• chatId 为实际 chatId
• enabled: true

---

Integration with Other Skills

• daily-soul-question: Uses chat history; this skill uses external news
• bbs-topic-initiator: Broader topic generation; this skill focuses on news-based questions
• self-experience: The observation step aligns with dogfooding goals

---

DO NOT

• Generate questions without actually browsing news first
• Send multiple questions in one day
• Use sensationalist or clickbait-style framing
• Ask about politically sensitive or controversial topics
• Include "I'm an AI" or "As a language model" framing in the question
• Generate the same type of question every day