blog-post for Claude Code: a skill that runs one blog post from sources to deployment

blog-post is a skill that takes one approved topic from source collection through deployment verification in a single run. This post covers the problem it solves, who it suits, how to set it up and trigger it, and the one thing that stops people when they try to reuse it.

What problem it solves

An agent can write a blog post in a minute, and that is the problem. Left alone, it writes from general knowledge, fills gaps with confident commentary, and ships something that reads like every other post on the topic. A blog that publishes weekly needs the opposite: posts built on what the author actually saw, checked the same way every time before they go live.

blog-post fixes the order. It first confirms the topic has been evaluated, then looks for real material such as existing drafts, page screenshots, verbatim reviews, and logos. If a draft exists, it becomes the primary material rather than something to rewrite from scratch. Next it interviews the author about where they got stuck, how their judgment changed, what to warn readers about, and how to read the numbers. Commentary with no answer behind it is left out. This is one of the four content marketing skills I use on my own blog, which is why it has no public repository.

Who it fits

The chapter rates four situations. Three are enough to decide.

Situation	Fit	Reason
Publishing one post a week on an active blog	Good fit	Applies the same gates each time
A topic with hands-on screenshots and reviews	Good fit	Passes the real-material gate
Posting a short announcement quickly	Poor fit	Long interview and confirmation stages

A case study that quotes reviews sits in the middle. It works, but you have to check the originals.

Setup and first run

There is no repository to install from. You build your own skill from the SKILL.md structure, with the gate script sitting next to it in the blog repository.

# Example layout from the blog repository root
my-blog/
  .claude/skills/blog-post/SKILL.md
  scripts/check-post.sh

It activates on requests such as “Write a blog post” or “Publish a post on this topic.” A topic that has not been evaluated goes through evaluation first.

Publish a post on this topic

From there the run follows three stages. It collects material and checks topic approval. It interviews the author and starts writing only after five items, including the title, target reader, and outline, are confirmed. Then it runs the gate script and a review agent in a separate context, deploys, and checks the live URL directly.

What trips people up

The skill is not self-contained. Its body paths, deployment commands, and domains are the author’s environment values, so they all need replacing for your blog. More importantly, the gate checks and the scripts for image compression and cover rendering live outside the skill folder. Copy only SKILL.md and those steps have nothing to call. Topic evaluation, logged-in screen captures, diagrams, and review also belong to other skills and agents, so plan to build or substitute those too. One pause point is worth keeping as is: posts with sensitive screens show the processed images to the user for confirmation before deployment.

Where this comes from

This post adapts part of chapter 02-04 of The Claude Code & Codex Skills Guidebook. The full chapter covers the separate unattended procedures for posts scheduled ahead of a holiday and announcement posts published the day after a product launch, and the stage-by-stage breakdown from personal information anonymization to cover specifications, and the book has 52 more skills like this one. Read the full chapter in The Claude Code & Codex Skills Guidebook on Gumroad.