Free resources / Agent skills
The blog writer
that reads human.
A Claude Code skill that turns your agent into a real ghostwriter. You give it a topic and your own take, and it does the rest: it researches the subject across the web and the communities where your audience actually talks, finds the angle the top results are missing, and proposes a plan for you to approve. Only then does it write the full SEO, AEO and GEO optimized article in your voice, generate the graphics, and hand you a piece that is ready to publish. No invented stats, and zero AI artifacts.
Real result
New domain. One article.
83 visits in 30 days.
This is not a polished case study with years of authority behind it. It was a brand new website on a brand new domain, with no audience, no email list and no backlinks pointing at it. A single article written with the skill reached 83 visits in its first 30 days, entirely from a standing start. No ad spend, no promotion, just one well-researched piece doing the work that one piece is supposed to do.
How it works
Research first, write second, never fabricate.
Eight steps, in order, with a hard stop for your approval before a single word is written. The skill does the research and the drafting, you keep every editorial decision. Here is exactly what happens from a topic to a finished article.
- 01
Gather
Before anything else, it reads what only you can give it: the topic, your point of view, and your real assets. Your own data, customer results, screenshots, numbers, the things you have lived through. Generic articles all sound the same because they have none of this. Your real material is the raw difference between a piece that could have come from anyone and one that could only have come from you, and the skill treats it as the backbone of the whole article.
- 02
Research
Next it researches the topic with whatever web tools are connected: search, page fetching, a browser. It pulls two things and two things only, facts and angles. It reads the competing articles for context and sources, and it reads Reddit, forums and communities to hear the real questions people ask, the complaints they have, and the exact language they use to describe the problem. None of that research ever becomes the voice of the article. It feeds the substance, never the style.
- 03
Information gain
It then studies what the top-ranking pages already cover and, more importantly, what they all leave out. Your article is built around that gap, so it adds something genuinely new to the conversation instead of being the tenth reworded copy of page one. This is the single biggest reason a piece earns its way up the rankings: search engines and answer engines both reward the source that says something the others did not.
- 04
Propose a plan
Before writing a word, it shows you the full plan: the outline section by section, what each section will cover, and the exact list of graphics it intends to create. Then it stops. You see the whole shape of the article up front, so there are no surprises and no thousand-word draft heading in the wrong direction.
- 05
Wait for approval
Nothing gets written until you say go. You can rework the outline, change the angle, swap the graphics, widen or narrow the scope, or send it back to research. This is your checkpoint and your steering wheel. The skill does the heavy lifting, but the editorial decisions stay entirely with you.
- 06
Write
On approval it writes the full article as you, in your voice, following the approved outline and a modern AEO-friendly skeleton. It holds the anti-AI rules the entire way through: no banned filler words, no em dashes, no en dashes, no smart quotes, no robotic transitions. The output is a finished, publish-ready draft, not a rough outline you still have to rewrite.
- 07
Verify, never fabricate
If it is not certain a figure or fact is real, it refuses to invent one. Instead it drops a clear VERIFY marker in the text so you can confirm the number or replace it with a real one. No hallucinated statistics, no made-up sources, no confident nonsense. What ships is either verified or visibly flagged for you to check.
- 08
Generate graphics
Finally it produces the graphics from the approved plan using whatever image provider is connected, so the piece arrives complete: words and visuals together, formatted and ready to publish. You go from a topic to a finished, illustrated article in one pass, then hand it straight to PostSider to schedule and post.
Why it works
Made to rank, made to be quoted.
Plenty of tools spit out a thousand generic words in seconds. That is the easy part, and it is also why most AI content never ranks and never gets read. This skill is built around the four things that actually move the needle in the age of search and AI answers.
It adds something new
The information-gain pass is the quiet reason this works. Because the article is built around what the top results are missing, it is not a reworded version of page one. It contributes something to the topic, and that is exactly what search engines reward with rankings and what readers reward with their time. Sameness is invisible. Information gain is what gets you seen.
Built to be quoted by AI
Search is no longer only ten blue links. People ask AI answer engines, and those engines lift their answers from pages shaped a particular way. The skill writes on an AEO-friendly skeleton on purpose: a TL;DR, thesis-style headings, dry-answer blocks that state the answer plainly, and an FAQ section with JSON-LD. That is the precise structure ChatGPT, Google AI and Perplexity pull from. You do not just get indexed, you get quoted.
It sounds like you
The fastest way to lose a reader is to sound like a machine. Because it writes in your voice and builds on your real assets, the article reads like a person actually wrote it, with a point of view and lived detail behind it. That human quality is what keeps people on the page, earns shares, and lines up with everything Google says it rewards about helpful, first-hand content.
Zero AI artifacts
No em dashes, no en dashes, no smart quotes, no "in conclusion", no "in today's fast-paced world", none of the tells that make a piece smell AI-generated. The anti-AI rules run the entire time it writes, not as a cleanup pass at the end. The result reads human because it is engineered to, line by line, so neither your readers nor a detector flag it as machine output.
Install
Two commands and a slash.
No setup, no config, no API keys to wire up. Clone the skill into your Claude Code
skills folder, start a new session, and /blog-writer is there as a slash
command. Give it your topic plus your own opinion and information, and it takes over
from research to finished article. Grab it with the button above and we email you the
link.
git clone https://github.com/lumizone/blog-writer-claude-skill.git
cp -R blog-writer-claude-skill/skills/* ~/.claude/skills/
# new session, then:
/blog-writer write a blog post about <your topic>Almost there
Taking you to the skill
We are opening this skill for you in a second. Want to hear when we ship new skills and updates? Drop your email and we will keep you posted. It is optional, and we never clutter your inbox with junk.