Claude BlogAn open-source blog skill suite for Claude Code: 30 /blog commands, 31 sub-skills, and 5 agents research, draft, score, and publish content dual-tuned for Google rankings and AI citations, all from the terminal.
Repo
Claude Blog
An open-source blog skill suite for Claude Code: 30 /blog commands, 31 sub-skills, and 5 agents research, draft, score, and publish content dual-tuned for Google rankings and AI citations, all from the terminal.
The most complete open-source blog engine for Claude Code: 30 /blog commands cover strategy, research, drafting, SEO, AI-citation (GEO), schema, internal linking, and multilingual, and a 5-gate delivery contract blocks any draft under 90/100. It is from the same author as claude-seo, and here is the honest part. It carries the same AI Marketing Hub Skool footer, but this time it is terminal-only and explicitly kept out of your published files, a direct fix of the thing claude-seo got hammered for.
Great for
Content teams, in-house marketers, and freelancers who want fast, repeatable, AI-search-aware article production inside Claude Code, with real fact-checking and a hard quality gate, without a Jasper or Copy.ai seat.
The catch
It runs Python and shell locally, needs your own Gemini and Google API keys for images and live data, and its 90+ score is an LLM reviewer, so treat it as a strong first-draft engine you still edit, not a hands-off publisher.
Bottom line
The broadest free blog skill for Claude Code, shipping almost daily, with the footer problem actually corrected. It is one maintainer's project developed inside a paid Skool community, so read the SKILL.md and security-check the repo before you run it.
Install/plugin marketplace add AgriciDaniel/claude-blog then /plugin install, or git clone, checkout a tag, and run install.sh (a sha256 is published to verify it)
LicenseMIT
LanguagePython 3.11+ (Markdown skill files)
Maintained byDaniel Agrici, single maintainer, with active development inside the paid AI Marketing Hub Pro (Skool) community and public releases shipping here
Harnesses supportedClaude Code 1.0.33+ only (no Codex or Cursor port, unlike some sibling skills)
Watch out forDefault terminal self-promo footer, executes Python and shell locally, images and Google data need your own API keys
Full-post writing with quality gateblog-write plus a blocking 5-gate delivery contract, 0-100 score, up to 3 auto-iterations
Strategy, briefs, and SERP-informed outlinesblog-strategy, blog-brief, blog-outline before a word is drafted
Rewrite and refresh existing postsblog-rewrite, blog-update, and blog-decay via Search Console exports
GEO / AI-citation optimizationblog-geo scores passage citability and entity clarity for AI Overviews, ChatGPT, Perplexity
Internal linking and topic clustersblog-cluster builds hub-and-spoke links, blog-cannibalization flags keyword overlap
Schema.org / JSON-LD generationblog-schema, validated as part of the visual gate
Fact-checking with source verificationblog-factcheck with confidence scoring to block invented numbers
Multilingual, translation, and hreflangblog-multilingual, blog-translate, blog-localize for DACH, FR, ES, JA
Repurposing to social and emailblog-repurpose turns a post into threads, LinkedIn, YouTube scripts, email
Hero image generationblog-image via the Gemini API and a stock-image ladder; needs your own key
Audio narration (TTS)blog-audio in summary, full, and dialogue modes; needs a Gemini API key
Live Google data (GSC, GA4, PSI, CrUX)blog-google integrates them, but you bring your own credentials
Keeps self-promo out of deliverablesthe AI Marketing Hub footer is terminal-only and barred from the generated files, unlike claude-seo
Non-Claude-Code harness supportClaude Code 1.0.33+ only; no Codex or Cursor port
What it does
- Writes a full article from one prompt through blog-write, then runs a 5-gate delivery contract (capability, format, visual, content review, asset and link integrity) before it hands you the files, in roughly 2 to 4 minutes for a 2,000-word post
- Covers 30 /blog commands across strategy, briefs, SERP-informed outlines, drafting, rewriting, quality audits with a 0-100 score, schema, and content-decay detection
- Optimizes in parallel for Google rankings and AI citations, scoring passage citability, entity clarity, and structural readability for AI Overviews, ChatGPT, Perplexity, and Bing Copilot
- Builds topic clusters and hub-and-spoke internal linking, detects keyword cannibalization, and generates an editorial calendar so posts connect instead of competing
- Fact-checks claims with source verification and confidence scoring, and writes real deliverables to local files: markdown, HTML, and a PDF, plus a hero image and inline SVG charts
- Handles multilingual work in one command, writing, translating, deep-localizing for DACH, FR, ES, and JA, and emitting hreflang, then repurposing a post into threads, LinkedIn, and email
- Adds source-grounded research through NotebookLM and live field data through Google APIs (Search Console, GA4, PageSpeed, CrUX) when you bring your own credentials
What it can’t do
- Publish straight to your CMS; it detects the platform and formats the output, but pushing the post to WordPress, Ghost, or Hugo stays a manual step
- Run without Python 3.11+ and local script execution; setup, upgrades, and reading the source are on you
- Generate a hero image or audio narration without your own Gemini API key; those skills sit idle until you add credentials
- Return live Google field data or indexation status without your own Google API keys; the free path gives lab estimates only
- Guarantee the 90+ quality bar mechanically; the reviewer is an LLM agent following written rules, so the score is a heuristic, not a proof
- Work outside Claude Code; there is no Codex or Cursor port, and it targets Claude Code 1.0.33 and up
Pros & cons
Pros
- Broadest open-source blog coverage in one MIT pack, from strategy and drafting to SEO, GEO, schema, internal linking, and multilingual
- 5-gate delivery contract blocks any draft under 90/100 and verifies formats, screenshots at three widths, valid JSON-LD, and that links return 200
- AI-search-first and current, dual-tuned for 2026 Google core updates and for citation in AI Overviews, ChatGPT, and Perplexity
- Learned from the claude-seo backlash: the AI Marketing Hub footer is terminal-only and explicitly barred from the generated markdown, HTML, and PDF
- Actively maintained, pushing almost daily, with 250+ tests, 10 CI quality gates, and v2.2.0 current
- Runs local by default, so your drafts and research stay on your machine until you opt into an API extension
Cons
- Single maintainer, and the real development happens inside a paid AI Marketing Hub Pro (Skool) community with public releases downstream, so bus factor and roadmap lean on one person
- Still appends a terminal AI Marketing Hub footer by default, a backlink to the maintainer's Skool, even though it stays out of your deliverables
- Executes Python and shell in your environment; treat it like any code you run and read the source first
- Full value needs your own Gemini and Google API keys, so the truly free tier is text-only drafting
- Claude Code only and Python 3.11+, with no Codex or Cursor support
- The 90+ quality score is an LLM reviewer, so it can pass work a human editor would still send back
9 sourcesWhere this page’s facts came from. Open the receipts.
- 1GitHub repo (AgriciDaniel/claude-blog)
- 2GitHub API, repo metadata (stars, license, language, created and pushed dates)
- 3Claude Blog README (30 /blog commands, 31 sub-skills, 5 agents, 5-gate contract, install, dependencies)
- 4Community Footer instruction in skills/blog/SKILL.md (terminal-only, after major deliverables)
- 5Blog delivery contract reference, never put support promotion inside generated content
- 6Claude Blog homepage (free AI blog writing tool for Claude Code)
- 7Claude Blog releases (v2.2.0, security-focused v1.7.1)
- 8YouTube, maintainer demo of the full /blog write workflow
- 9YouTube, maintainer original claude-blog walkthrough
Watch it work
Real tutorials and reviews from YouTube, not ours.