blog-topic-judge for Claude Code: a skill that decides whether a topic is worth a post

blog-topic-judge is a skill that evaluates a blog topic on five axes and decides whether it should be written at all. This post covers the problem it solves, who it suits, how to set it up and trigger it, and the one thing that catches people when they adapt it.

What problem it solves

Most blogs do not fail because posts are badly written. They fail because the topics drift. A video script gets turned into a post because it already exists, a trending subject gets covered because it is trending, and after a few months the blog is a mixed bag with no clear reader and no path to anything the author sells. Asking an agent “should I write this?” usually gets an encouraging yes.

blog-topic-judge makes the answer stricter. It divides the reader’s learning journey into four slots: understanding, exploration, comparison, and recommendations. A topic that fits none of them is rejected. A topic that fits but has no real material behind it, because the author has not tried the subject firsthand, gets conditional approval with that hands-on work as a prerequisite. 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
Deciding whether to turn a video script into a post	Good fit	A use case in the description
Filtering several candidate topics	Good fit	Evaluates each independently
A diary-like blog without a purpose	Poor fit	Assumes readers and a conversion path

Moving old blog posts sits in the middle. It works one post at a time, but bulk migration is prohibited.

Setup and first run

There is no repository to install from. You build your own skill with this layout in your blog repository, including a purpose document and a script that fetches engagement signals.

# Example layout from the blog repository root
my-blog/
  .claude/skills/blog-topic-judge/SKILL.md
  .claude/skills/blog-topic-judge/assets/decision-tree.html
  docs/blog-purpose.md
  scripts/topic-signal.sh

It activates on requests such as “Should I blog about this?” or “Evaluate this topic,” and it makes a judgment without writing anything.

Should I blog about this topic?

The run follows a fixed order. It checks reader and cluster fit first, then the conversion path, format, and whether there is enough real material such as captures, tables, and judgments. Last comes reuse, meaning whether the topic can feed other channels. Reuse is secondary and can only rank approved topics, never reverse a decision. The output is one of three verdicts with reasons by axis, plus a single rework proposal that combines a title, section structure, and ending call to action. On borderline topics it takes the stricter option.

What trips people up

High engagement on social media is tempting to read as a green light, and the skill is written to resist that. The instructions say to check which channel the numbers come from and what the blog’s current bottleneck is before letting them set production priority, and not to claim causation without a control group. The catch is that the script that retrieves those engagement signals lives outside the skill folder. SKILL.md tells the agent to fetch signals with that command rather than estimate them, so if you copy only the skill, that step has nothing to call. Build the script, or decide deliberately to drop the signal step.

Where this comes from

This post adapts part of chapter 02-05 of The Claude Code & Codex Skills Guidebook. The full chapter covers the complete six-row table of evaluation axes with the question each one asks, and how to replace the blog purpose, reader personas, and CTA products that are set to the author’s environment, and the book has 52 more skills like this one. Read the full chapter in The Claude Code & Codex Skills Guidebook on Gumroad.