starchild-ai-agent/@1892/agent-blog

Agent Blog Skill

Manage a blog built on the blog template. Full lifecycle: write or import drafts, format articles, set titles and hero images, add SEO structures, build, preview, and publish.

Project layout

config.py                  site constants (SITE_URL, SITE_NAME, CTA, UTM, OG colors)
build.py                   generator; python3 build.py [--check|--drafts]
scripts/import_article.py  import any HTML article as a draft
scripts/import_x_articles.py  import X articles as drafts
scripts/og_image.py        auto-generate OG cards from post titles
templates/                 base.html, home.html, post.html, tag.html
assets/                    style.css, fonts, icons, vendor/chart.umd.min.js
assets/images/             hero images live here
content/posts/             YYYY-MM-DD-<slug>.md source files
public/                    build output (gitignored)

Config

config.py is the single source of truth. If it does not exist, the build falls back to config.example.py defaults (with a printed note), so a fresh fork builds with zero setup. To customize: copy config.example.py to config.py and edit the values.

Key fields: SITE_URL, SITE_NAME, SITE_TAGLINE, CTA_URL, CTA_TEXT, OG_* colors, UTM_SOURCE.

Writing a post

Create content/posts/YYYY-MM-DD-<slug>.md:

---
title: "Post Title"
slug: post-title
date: 2026-01-15
author: Author Name
description: "Short SEO summary."
tags: [tag1, tag2]
draft: true
og_image: /assets/images/hero.png
---

Then python3 build.py. Drafts render to public/drafts/<slug>/ and are excluded from homepage, RSS, sitemap. Set draft: false to publish.

Frontmatter fields

| Field | Required | Notes | |-------|----------|-------| | title | yes | Wrap in quotes if it contains : or # | | slug | no | Defaults to filename slug | | date | no | Defaults to filename date | | author | no | Defaults to config.DEFAULT_AUTHOR | | description | no | Falls back to first 160 chars of body | | tags | no | YAML list [a, b] | | draft | no | true = excluded from public surfaces | | og_image | no | Custom OG image path; auto-generated if omitted | | source_url | no | Original URL if imported |

Body

Standard markdown after the ---. Supports headings, bold, italic, links, images, code blocks, blockquotes, lists, tables, horizontal rules.

For raw HTML articles (imported), the body can contain inline <style> and <script> blocks. The importer scopes CSS under .sr-article.

Importing articles

From any HTML URL

python3 scripts/import_article.py "URL" \
    --date YYYY-MM-DD --slug S --title "T" \
    --author "Author" --tags a,b --description "..."

The importer fetches the page, extracts body + styles + scripts, scopes CSS under .sr-article, drops the original <h1>, rewrites Chart.js CDN to vendored copy, and writes a draft markdown file.

From X articles

python3 scripts/import_x_articles.py

Reads twitter_articles_5_raw.json (or similar export), downloads cover images, and writes formatted markdown posts.

Building

python3 build.py --check    # validate only, write nothing
python3 build.py            # full build (drafts excluded from public surfaces)
python3 build.py --drafts   # full build (drafts visible on homepage, marked [draft])

Output goes to public/. The build is deterministic.

OG images

If a post has no og_image in frontmatter, scripts/og_image.py generates a 1200x630 PNG card with the post title, date, and brand wordmark. Colors come from config.py (OG_* fields). A default card is also generated for the site homepage.

UTM analytics

Every outbound link (nav CTA, footer, article CTA) is tagged at build time:

• utm_source = config.UTM_SOURCE (default: blog)
• utm_medium = nav, footer, or article_cta
• utm_content = post slug (only on article CTAs)
• utm_campaign = config.UTM_CAMPAIGN (optional)

Internal links, assets, RSS, and JSON-LD are never tagged. This lets analytics platforms attribute outbound clicks to specific articles without any client-side setup.

Previewing

Serve public/ with any static server:

python3 -m http.server 8000 -d public/

On Starchild, use the preview system to serve the built site and get a shareable URL.

Publishing

The public/ directory is static HTML. Deploy to any static host:

• GitHub Pages
• Netlify / Vercel
• Any static file host (rsync, SCP)
• Starchild community publish (via community-publish skill)

Validation checklist

Before publishing a post:

1. python3 build.py --check passes with no errors
2. Post has a title and description in frontmatter
3. draft: false is set
4. Hero image exists at the path specified in og_image (if set)
5. python3 build.py completes successfully
6. Preview the post at /posts/<slug>/ to verify rendering
7. Check /feed.xml and /sitemap.xml include the post

Troubleshooting

• YAML parse error: usually an unquoted title containing : or #. Wrap in double quotes.
• Charts blank: a CDN <script src> slipped back in. Confirm the built page references vendor/chart.umd.min.js.
• Wrong fonts on imported article: re-run the importer (it forces white bg + blog fonts). Do not patch public/ by hand.
• Mojibake: re-run the importer; it decodes UTF-8 via apparent_encoding.
• "config.py not found" note during build: not an error; the build used example defaults. Run cp config.example.py config.py and edit to customize.
• Telegram / X link preview shows no image despite og:image being set: hero image was saved as RGBA PNG. Telegram's crawler silently rejects alpha-channel PNGs with no fallback. Fix: Image.open(path).convert('RGB').save(path, format='PNG'). Verify: struct.unpack('B', open(path,'rb').read()[25:26])[0] == 2 (color type 2 = RGB). Also add og:image:secure_url and og:image:type meta tags for full Telegram/X card support.
• On Starchild, use preview(action="serve") to get a shareable /preview/<id>/ URL — never expose localhost:<port> to users. Web users see the Preview panel; on Telegram/WeChat use the community-publish skill's publish_preview() for a public URL. Protected slugs (1892-starchild-blog, 1892-docs) must never be stopped or overwritten.