Skip to content
Education
Skill

/industry-news-insight

Generate a concise Chinese WeChat-ready Markdown article from an industry name by searching current news from the last month, verifying evidence across production, demand, inputs, ecosystem, policy, financing, and technology signals, and writing first-person neutral third-party

BOOST
From plugin
blog
8.6k85 skills
Install
$ npx -y skills add digoal/blog --skill industry-news-insight --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.
  • Slash command/industry-news-insight

Context preview

The summary Claude sees to decide when to auto-load this skill.

Generate a concise Chinese WeChat-ready Markdown article from an industry name by searching current news from the last month, verifying evidence across production, demand, inputs, ecosystem, policy, financing, and technology signals, and writing first-person neutral third-party

SKILL.md

industry-news-insight.SKILL.md
name: industry-news-insight
description: Generate a concise Chinese WeChat-ready Markdown article from an industry name by searching current news from the last month, verifying evidence across production, demand, inputs, ecosystem, policy, financing, and technology signals, and writing first-person neutral third-party industry insight without visible author identity or AI traces. Use when the user asks for 行业新闻追踪, 行业洞察, 近一个月行业分析, 公众号短文, industry trend notes, or a source-backed short article saved to the current project's markdown directory.

Industry News Insight

Overview

Use this skill to turn one industry name into a short, evidence-backed Chinese Markdown article for WeChat-style mobile reading. The output should read like a human industry observer's first-person note, not a formal report, pitch, consultant bio, or AI-generated summary.

Workflow

1. Confirm the target industry and date window.

  • Default to the most recent 30 days from the current date.
  • If the industry name is broad, keep it broad unless the user specifies a sub-sector; avoid silently narrowing the scope.
  • State the date window in the working notes or final response, but do not over-explain it inside the article.

2. Search current evidence.

  • Search web/news sources for the industry name plus terms such as 新闻, 政策, 投融资, 价格, 产能, 需求, 出口, 技术, 并购, supply, demand, market, regulation, funding, earnings, capacity, and price.
  • Prioritize sources published within the last 30 days. Use older sources only for background or to explain a structural fact, and label them as background in notes.
  • Prefer primary or high-signal sources: regulators, exchanges, statistics agencies, company announcements, earnings materials, industry associations, reputable business media, and credible research institutions.
  • Collect enough evidence to cover at least three angles among production端, 消费端, 生产资料/上游投入, 渠道/生态, 政策监管, 技术变化, 资本市场/投融资, and international trade.

3. Validate the story before writing.

  • Separate facts from interpretation.
  • Look for contradictions across sources; resolve them by source quality, date, and measurement definition.
  • Do not write if the last-month evidence is too thin, stale, promotional, or one-sided. Tell the user what is missing instead of producing a weak article.
  • Add supplemental searches when a key causal link depends on missing evidence.

4. Build the article thesis.

  • Use a "news signal -> industry mechanism -> what may change next" structure.
  • Avoid a simple news roundup. Extract one core judgment from the evidence.
  • Make the judgment neutral and falsifiable; avoid investment advice, buy/sell calls, and unsupported predictions.

5. Write for mobile WeChat reading.

  • Write in Chinese unless the user asks otherwise.
  • Keep the article under four mobile screens: normally 900-1500 Chinese characters, excluding references and diagrams.
  • Use short paragraphs, compact headings, and direct sentences.
  • Use first person naturally, such as "我更关注的是..." or "我把这件事放到产业链里看...".
  • Do not reveal an author role. Avoid phrases like "作为第三方顾问", "作为AI", "本文由AI生成", "我们团队", or "研究报告认为".
  • Do not use obvious AI template language such as "综上所述", "值得注意的是" repeated mechanically, "未来可期", or inflated slogans.
  • Do not fabricate quotes, statistics, rankings, or company positions.

6. Add visuals only when they improve comprehension.

  • Prefer one compact Mermaid, SVG, or ASCII diagram when it clarifies a value chain, feedback loop, or supply-demand relation.
  • Keep visuals mobile-readable and text-light.
  • Skip visuals when they would consume space without adding insight.

7. Save the Markdown.

  • Create or use the current project's `markdown/` directory.
  • Save as `markdown/<YYYYMMDD>-<industry-slug>-industry-news-insight.md`.
  • Include the article title, body, optional diagram, and a compact `参考来源` section with source names, dates, and links.
  • In the final response, report the saved absolute path and a brief evidence summary.

Article Shape

Use this structure unless the user requests another format:

# <短标题:行业名 + 核心判断>

<开场:最近一个月最关键的新闻信号,用一两段进入,不写套话。>

## <机制标题>

<解释生产端、消费端、上游或生态中的关键变化。>

## <真正的变化>

<给出第一人称观察和行业机制判断。>

<可选 Mermaid/SVG/ASCII 图>

## <接下来盯什么>

<列出2-4个后续观察指标,不写投资建议。>

---

参考来源:
- <来源名>,<发布日期>,<链接>

Quality Bar

  • The article must be readable as a standalone public-account post.
  • Every specific data point or event must be traceable to a source.
  • The core claim must follow from the cited evidence, not from generic industry common sense.
  • The article should feel concise, observational, and human; remove bureaucratic framing, consultant disclaimers, and AI-style transitions before saving.
Read more
Ships withblog

AI,Opensource,Database,Business,Finance,Minds. git clone --depth 1 https://github.com/digoal/blog

Get the whole plugin

Other skills on blog.

a-stock-prediction
Skill

a-stock-prediction

A股个股次日(T+1)走势预测 skill。当用户输入A股股票代码(如 600519、000858、300750、688981、430090)或公司名称(如"贵州茅台"、"宁德时代"、"比亚迪"),并要求预测次日走势、给出交易计划、评估盈亏概率时,触发本…

@digoal@digoalInstall Skill
andres-freund-perspective
Skill

andres-freund-perspect…

PostgreSQL 核心 committer Andres Freund 的思维框架与表达方式。基于 6 个调研维度(著作、对话、表达 DNA、他者视角、决策、时间线) 共 2957 行 / 200 KB 一手资料的深度调研,提炼 6 个核心心智模型、10…

@digoal@digoalInstall Skill
article-fact-checker
Skill

article-fact-checker

三层审查模型,逐段逐句验证文章真伪、证据链与逻辑结构。Use when the user asks to fact-check, verify, audit, or evaluate the credibility of an article, essay,…

@digoal@digoalInstall Skill
axiom-explainer
Skill

axiom-explainer

Write Chinese Markdown articles for university students and adults with social experience that explain one viewpoint, axiom,…

@digoal@digoalInstall Skill
blog-cover-poster
Skill

blog-cover-poster

Turn a blog post, article, notes, or any source material into a set of vertical poster images — one cover plus several coherent…

@digoal@digoalInstall Skill