blog-publisher

name

blog-publisher

description

Publish bilingual Astro blog posts for blog.mushroom.cv and create WeChat Official Account drafts. Trigger when the user says: 发布, 发布文章, 发布blog, 发布公众号, publish blog, create WeChat draft. Use for: - turning user-provided notes, links, README summaries, or markdown into a blog post - generating or processing a 1200x630 banner - publishing to Cloudflare Pages - creating a WeChat draft from the final markdown Critical rules: - read this skill before publishing - filenames must be English slugs, never Chinese - default category is Tech-News - run SEO/GEO checks before deploy - finalize the slug before Blog deploy and before WeChat draft creation

Blog Publisher Skill

Mission

Publish one complete bilingual article to the Astro blog and, when requested, create a WeChat Official Account draft from the same final markdown.

Primary site:

• Blog: https://blog.mushroom.cv/blog/SLUG/
• WeChat backend: https://mp.weixin.qq.com

Project paths:

• Blog articles: src/content/blog/
• Images: src/assets/images/
• M1 Blog publisher: pipeline/m1/publisher.py
• M2 WeChat publisher: pipeline/m2/index.js
• M3 references, if needed: pipeline/m3/
• Blog deploy: pnpm build + wrangler pages deploy
• WeChat draft output: pipeline/m2/output/

Trigger Words

Use this skill immediately when the user says any of:

• 发布
• 发布文章
• 发布blog
• 发布公众号
• 发布：文件路径.md
• publish blog
• create WeChat draft

Required Output

Return:

• final Blog URL
• WeChat draft status and media ID, if M2 was requested
• final markdown path
• final banner path and size
• any validation issue that remains

Non-Negotiable Rules

1. Article filenames must be English slugs only.
2. Do not use Chinese characters in markdown or image filenames.
3. Every article frontmatter must include category.
4. Dates use ISO format: YYYY-MM-DD.
5. Default category: Tech-News.
6. New articles should appear at the top of /blog/.
7. Add the required copyright block to both Chinese and English sections.
8. Run SEO/GEO checks before deploy.
9. Create the WeChat draft only after the final slug and Blog URL are confirmed.
10. Do not create duplicate routes for the same article.

Allowed categories:

• Tech-News default
• Tech-Experiment
• Progress-Report
• Research
• DN

Standard Frontmatter

---
title: "中文标题"
titleEn: "English Title"
description: "中文描述"
descriptionEn: "English description"
pubDate: "YYYY-MM-DD"
updatedDate: "YYYY-MM-DD"
category: "Tech-News"
tags: ["tag1", "tag2", "tag3"]
heroImage: "../../assets/images/SLUG-banner.jpg"
---

Required Copyright Blocks

Chinese section, before <!--EN-->:

---

> © 2026 Author: Mycelium Protocol. 本文采用 [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/deed.zh) 授权——欢迎转载和引用，须注明作者姓名及原文链接，不得去除署名后以原创发布。

English section, at the end:

---

> © 2026 Author: Mycelium Protocol. Licensed under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) — free to share and adapt with attribution. You must credit the author and link to the original; removing attribution and republishing as original is not permitted.

Fast Path: Existing Topic Or Short Article

Use this path when the user provides the topic, source link, README/project info, or a short post draft.

Target time: 3-5 minutes, excluding external API/network delays.

1. Decide The Final Slug First

Choose a short English slug before writing files.

Good:

• ai-ppt-master-native-powerpoint.md
• moss-tts-nano-local-tts.md

Bad:

• Chinese filenames
• long title-derived filenames
• truncated filenames generated by tools

If another article already has today’s pubDate, choose a slug that will sort early enough to place the new post near the top. Current blog list sorting is primarily by pubDate; same-day ordering can depend on collection/file order.

2. Prepare Banner

If the user provides no image, either:

• generate a custom banner when requested, or
• choose one default banner from src/assets/.

For custom banners:

convert INPUT.png -resize 1200x630^ -gravity center -extent 1200x630 -quality 85 src/assets/images/SLUG-banner.jpg
ls -lh src/assets/images/SLUG-banner.jpg

Preferred constraints:

• size: 1200x630
• file size: under 120KB when practical
• filename: English only

Default banner pool:

• blog-placeholder-1.jpg
• blog-placeholder-2.jpg
• blog-placeholder-3.jpg
• blog-placeholder-4.jpg
• blog-placeholder-5.jpg
• banner-human-ai-coexistence.jpg
• banner-cypherpunk-revolution.jpg
• banner-mycelial-network.jpg
• banner-future-is-now.jpg
• banner-ai-new-intelligence.jpg
• banner-digital-public-goods.jpg
• banner-ai-smart-city-collab.jpg
• banner-org-ai-transformation.jpg
• banner-ai-city-ecosystem.jpg
• banner-ai-personal-assistant.jpg
• banner-personal-growth-ai-skills.jpg

Default banner frontmatter format:

heroImage: "../../assets/banner-digital-public-goods.jpg"

Custom banner frontmatter format:

heroImage: "../../assets/images/SLUG-banner.jpg"

3. Create Final Markdown Directly

For fast publishing, write the final file directly:

src/content/blog/SLUG.md

Do not use pipeline/m1/publisher.py to save the article when the final slug matters. It derives a filename from titleEn, truncates long names, and can create duplicate or undesirable routes.

4. Run SEO/GEO Check

Read and apply .agents/skills/seo-geo/SKILL.md.

Minimum checks:

• description and descriptionEn exist
• tags has at least 3 items
• updatedDate is set
• slug is English and keyword-bearing
• opening has BLUF
• at least one H2/H3 is a question
• article includes concrete numbers or source facts
• at least one authoritative external link is present when source-based
• copyright blocks include Mycelium Protocol

5. Build Once

pnpm build

Check:

ls dist/blog | grep SLUG

6. Deploy Once

npx wrangler pages deploy dist --project-name=blog-mushroom --branch=main --commit-dirty=true

7. Validate Blog

curl -I https://blog.mushroom.cv/blog/SLUG/
curl -Ls --compressed https://blog.mushroom.cv/blog/ | rg -o '/blog/[a-z0-9-]+/' | head -5

Expected:

• article URL returns 200
• new article appears at or near the top
• no duplicate route exists for the old slug

8. Create WeChat Draft Last

Only run this after the final Blog slug and URL are confirmed.

cd pipeline/m2
node index.js "../../src/content/blog/SLUG.md" --theme blue

Themes:

• claude
• chengyun
• blue
• sticker
• mint
• purple
• cyber
• rose

Return the media ID from pipeline/m2/output/SLUG.json.

Standard Path: Preexisting Markdown Or M1 Required

Use M1 only when the user explicitly gives a markdown file and accepts M1’s save/copy behavior, or when the workflow requires its automatic handling.

python3 pipeline/m1/publisher.py PATH/TO/ARTICLE.md --images PATH/TO/IMAGE.jpg

Important:

• M1 saves a new article file under src/content/blog/.
• If a same-slug file already exists, M1 appends a timestamp.
• M1 also runs Blog build and Cloudflare deploy unless skip flags are used.
• Avoid creating a same-slug article manually before M1.
• After M1, inspect the generated path before running M2.

Use M1 when:

• the user provides a markdown file and accepts automatic naming
• the existing pipeline should handle image copying and full Blog deployment
• preserving the historical P1/M1 publication flow matters more than exact slug control

WeChat-Only Path

Use when the Blog article already exists and the user only asks for a WeChat draft:

cd pipeline/m2
node index.js "../../src/content/blog/SLUG.md" --theme blue

If the command fails with proxy or network errors, rerun with appropriate network permission. If it fails with WeChat IP whitelist error 40164, the user must add the reported IP in the WeChat platform allowlist.

Available Repo Scripts

Use these only when they fit the current task:

Script	Purpose
./publish.sh content.txt	full publish flow, may require AI polishing
./publish-fast.sh content.txt	fast direct publish flow
./deploy.sh	deploy only

SEO/GEO Integration

When publishing, run a practical SEO/GEO pass rather than a long report unless the user asks for one.

Required article traits:

• bilingual content separated by <!--EN-->
• BLUF at the start of Chinese and English sections
• one question-style heading
• FAQ for articles over 1000 words, optional for shorter posts
• original/source data when available
• source link at the end for GitHub/project/news posts
• Mycelium Protocol copyright block

Validation Checklist

Before deploy:

• markdown filename is English only
• image filename is English only
• titleEn exists
• descriptionEn exists
• pubDate is today for a new article
• updatedDate is set
• category is set
• tags has at least 3 items
• heroImage points to an existing file
• copyright blocks exist in both languages
• pnpm build succeeds

After deploy:

• final Blog URL returns 200
• /blog/ list shows the new article at or near the top
• no old slug remains as an independent article
• WeChat draft is created from the final slug

Cleanup Rules

If a slug changes before final release:

• delete old local markdown, if any
• delete old M2 outputs: pipeline/m2/output/OLD-SLUG.html and .json
• rebuild and redeploy
• if the old production URL still returns 200, add a redirect:

/blog/OLD-SLUG/ /blog/NEW-SLUG/ 301

Place redirects in:

public/_redirects

Then rerun:

pnpm build
npx wrangler pages deploy dist --project-name=blog-mushroom --branch=main --commit-dirty=true

Known Failure Cases

Duplicate routes from M1

Cause:

• manually creating src/content/blog/SLUG.md
• then running M1, which creates another timestamped file

Prevention:

• use the fast path and do not call M1, or
• let M1 create the file and inspect the generated filename before continuing

M1-generated slug is truncated

Cause:

• M1 derives filename from titleEn and truncates long titles

Prevention:

• fast path: create src/content/blog/FINAL-SLUG.md directly
• run pnpm build and deploy manually

WeChat draft created before final slug

Cause:

• M2 was run before the Blog slug was finalized

Prevention:

• run M2 only after final Blog URL is validated

Cleanup:

• ignore or delete old local M2 output
• create a new draft from the final markdown

Article not first on list

Cause:

• same-day posts share the same pubDate
• list sorting uses pubDate, so same-day ordering may depend on collection/file order

Prevention:

• choose a final slug before deploy
• verify list order before M2
• avoid post-deploy slug changes

WeChat IP whitelist error

Error:

invalid ip x.x.x.x, not in whitelist

Fix:

• user logs into WeChat Official Account platform
• add the reported IP to the API whitelist
• retry M2 after about 1 minute

Protected Git branch

Blog deployment uses Cloudflare Pages deploy and does not require direct push to main. If committing code is needed, use a feature branch and PR.