ai-wechat-hotspot-writer
AI WeChat Hotspot Writer
Overview
Use this skill to turn recent AI news and social heat into a publishable WeChat or Feishu-style article package.
Default output:
- an
AI Daily News-style digest by default - a ranked, source-backed AI industry list
- a product radar section for Product Hunt and GitHub Trending
- optional Chinese WeChat long-form expansion
- cover or long-image prompts for GPT Image 2 and Nano Banana 2
- saved Markdown files when working in a repository
Workflow
- Resolve the time window and article intent.
- Choose the output mode:
daily_newsby default,wechat_longformonly when requested. - Gather AI hotspots from multiple source channels.
- Group duplicates and verify source reliability.
- Score heat, credibility, writing value, and visual potential.
- Draft the digest or article using references/article-template.md; for the referenced AI Daily News format, also follow references/daily-news-style.md.
- Generate image prompts using references/image-prompt-template.md.
- Check quality with references/output-checklist.md.
- Save outputs with
scripts/save_article.pywhen a workspace is available.
Resolve The Brief
Default language: Simplified Chinese.
Default article style: daily_news, modeled as a compact AI news feed rather than a long essay.
Use wechat_longform only when the user asks for a "公众号长文", "深度解读", "文章草稿", or a complete narrative essay.
Use daily_news when the user asks to match the referenced AI Daily News style, write a daily or weekly digest, or summarize hotspots quickly.
Default time windows:
- "today", "latest", "daily": last 24 hours
- "recent days", "最近几天": last 3 calendar days
- "this week", "最近一周": last 7 calendar days
- "recent weeks", "最近几周": last 14 calendar days unless the user specifies otherwise
- "monthly" or "last few weeks": last 28 calendar days
Always state the exact date range in the output. If the current date is required, verify it from the environment and use absolute dates.
Gather Hotspots
Read references/source-playbook.md before collecting.
Cover at least four source families when available:
- official AI company and product announcements
- credible tech and business media
- product and developer trend sources
- social or community discussion sources
If sensight is available, use it as the accelerator for AI social pulse, semantic social search, article retrieval, and model sentiment. If it is not available, browse manually.
For OpenAI, Anthropic, Google, Microsoft, Meta, xAI, DeepSeek, Alibaba, Tencent, ByteDance, and other fast-moving AI companies, verify important launch or product claims against primary sources when possible.
Rank And Select
Read references/scoring-rules.md before ranking.
Prefer:
- 5 to 8 core hotspots for a normal article
- 8 to 12 items for a weekly roundup
- 3 to 5 "most worth writing" items when the user wants topic selection
Each kept item must have:
- what happened
- why it is hot
- source list with links or source names
- confidence level
- WeChat writing angle
- image-prompt potential
Do not inflate weak rumors. Put unconfirmed but socially hot items into "观察中" rather than the core factual section.
Write The Article
Use references/article-template.md.
For daily_news, required sections:
- title:
AI Daily News - optional note block:
☀️ 说明 - date or weekly range heading
AI行业动态- optional image slot after the section heading
- numbered news items, each with one title and one compact paragraph ending with a source link label
今天值得关注的产品Product Hunt Top 5GitHub Trending Top 5
For wechat_longform, also include:
- title options
- AI速览
- main article
- source and heat table
- 可配图热点
- image prompt package
Keep the writing readable:
- explain abbreviations on first use
- avoid empty grand claims
- explain why the reader should care
- distinguish fact, interpretation, and speculation
Generate Long-Image Prompts
Use references/image-prompt-template.md.
Create two prompt types:
daily_cover_image: one compact image that can sit belowAI行业动态long_infographic: a vertical long image summarizing 3 to 7 hotspotswechat_cover_or_lead_image: one strong cover for long-form public-account articles
Generate prompts for both GPT Image 2 and Nano Banana 2 unless the user asks for only one model.
Do not generate images unless the user explicitly asks to run an image-generation skill or tool. This skill primarily outputs prompts.
Save Outputs
When working in a repository, save the final package under:
docs/ai-wechat-hotspot-writer/
Prefer the helper script:
python local-skills/ai-wechat-hotspot-writer/scripts/save_article.py `
--date YYYY-MM-DD `
--slug ai-weekly-hotspots `
--article-file <article.md> `
--prompts-file <image-prompts.md> `
--output-root .
The script writes:
docs/ai-wechat-hotspot-writer/YYYY-MM-DD-<slug>.article.mddocs/ai-wechat-hotspot-writer/YYYY-MM-DD-<slug>.image-prompts.md
If saving is not useful, return the article and prompts directly in chat.
Example Requests
- "用最近 3 天 AI 热点写一篇公众号文章,并给我长图提示词。"
- "整理最近两周 AI 行业动态,按热度排序,做成公众号选题。"
- "参考 AI Daily News 的形式,写一篇本周 AI 产品和行业热点公众号。"
- "把这些热点变成适合 GPT Image 2 / Nano Banana 2 的长配图提示词。"
- "给我一篇 AI 周报公众号,附来源和配图 prompt。"
- "按 AI Daily News 的样式生成今天的 AI 行业动态和产品榜。"
More from zhangga/aihub
yahoo-data-fetcher
Fetch real-time stock quotes from Yahoo Finance.
20xai-stock-sentiment
Real-time stock sentiment analysis using Twitter/X data via Grok. Use when analyzing stock ticker sentiment, tracking retail investor mood, or gauging market reaction to events.
17remotion
Remotion renderer for json-render that turns JSON timeline specs into videos. Use when working with @json-render/remotion, building video compositions from JSON, creating video catalogs, or rendering AI-generated video timelines.
15skill-hub-builder
Scaffold and maintain a personal skill hub repository for collecting, syncing, and distributing commonly used agent skills. Use when setting up a reusable skill hub, adding external or local skills, updating bundles, or preparing one-click install flows for a team.
12sensight
触发:查热搜/热榜(抖音/微博/小红书/推特/头条/百度);社交媒体语义搜索(微博/微信/小红书/推特,任意话题);查特定作者/账号近期发文动态;查 AI 行业资讯(论文/博客/模型发布/口碑/深度摘要)。不触发:通用知识问答、代码生成、静态事实类问题。
10long-run-harness
Use for long-running, multi-step work where the agent should keep driving toward the true final outcome across many cycles instead of stopping after one pass. Trigger when the user wants autonomous progress, durable state, session recovery, single-task execution, validation, and review gates in an existing git repository or a similarly structured workspace.
10