Building an AI Blog Editor with Claude Skills

I've been blogging for years, but I've never had an editor. My process has always been: write a draft, sleep on it for a few days, then come back with "fresh eyes" to do my own editing. It works, but it's not optimal. When I worked at a company with an engineering blog, we had a group of writers who would take turns editing each other's posts. Having another person's feedback and perspective was invaluable. But on my personal blog? I could ask a friend or colleague, but I don't like to bother people with that kind of commitment.

The Code Review Parallel

In software development, no matter how confident you are in your code, you always have another developer review it before merging. That fresh perspective catches issues you'd never notice yourself. It's not about skill level, it's about being too close to the problem.

Blog posts are the same way. After spending hours crafting an article, you lose objectivity. You know what you meant to say, so you read what you intended rather than what you actually wrote.

Which brings me to an intriguing possibility: what if AI could provide that objective perspective?

Enter Claude Skills

I've been using Claude Code for development work and found it significantly better than other AI coding assistants. So when I learned about Claude Skills, I wondered: could someone have built a skill to make Claude function as a technical blog post editor? I searched and found several skills for blog post writing, but nothing for editing. So I decided to build one myself.

Building the Skill

In true AI-assisted fashion, I asked Claude to develop its own editing skill. I pointed it to the official documentation, including the Complete Guide to Building Skills for Claude, and explained what I needed.

My requirements for a technical blog post editor:

Detect common writing issues:

• Long run-on sentences
• Overly verbose explanations
• Too many superlatives
• Repetition of concepts
• Poor logical flow
• Unexplained acronyms or technical terms
• Posts that are too long (should they be split into a series?)

Check technical accuracy:

• Validate code snippets for the specified language
• Verify that technical explanations are accurate

Provide organized, actionable output:

• Suggest concrete rewrites instead of just pointing out issues
• Generate editorial review in scratch/{post-slug}-editorial-review.md (I keep a scratch directory in all my projects for AI output)

Claude came up with additional checks I hadn't even considered, such as:

• Does the introduction hook the reader?
• Preferring active over passive voice
• Reviewing elements in the frontmatter for SEO
• Code to prose ratio - is there good balance?
• Verify all link text is descriptive
• Are asides adding value or distracting?
• Is the value that the reader gains clear?
• Are the examples relatable?
• Is the conclusion satisfying?
• Categorizing recommendations by priority: Must Address, Should Address, and Nice to Have

Using the Skill

Invoking the skill is simple—use the /blog-editor command (Claude makes skills available via slash-command syntax) with the path to a draft blog post. For example:

/blog-editor src/markdown/building-a-no-frills-meditation-app.md

The skill reads the post, and generates a detailed editorial review with specific suggestions in the scratch directory.

Real-World Results

I've now used this skill for several blog posts, and it feels remarkably close to having another human review my work. It's caught numerous issues I would have missed:

• Repetition tracking: In long posts, I sometimes repeat a concept multiple times without realizing it. The skill catches this every time.
• Logical flow: It's helped reorganize sections to improve the narrative structure.
• Sentence tightening: It excels at identifying overly long sentences and suggesting more concise alternatives.
• Transitions: It points out where transitions between sections are needed to avoid jarring jumps in topic.

Here's a sampling of feedback from the review of an upcoming post about building vanilla JavaScript routing (the full review was much more comprehensive):

**Word Count**: 3,809 words
**Reading Time**: ~15-19 minutes

**Verbosity (Opening paragraph)**
Original: "There's something deeply appealing about vanilla JavaScript. In a world of constantly changing frameworks and build tools, I find myself drawn to the simplicity of writing code that just works with what the browser provides. No complex build systems with countless dependencies that become fragile as Node.js versions update and tooling incompatibilities emerge..."

Suggested: "There's something deeply appealing about vanilla JavaScript. In a world of constantly changing frameworks and build tools, I'm drawn to code that just works with what browsers provide—no fragile dependency chains, breaking webpack configs, or afternoon-consuming package-lock conflicts."

(Cut ~30 words while maintaining impact)

**Repetition Tracking**
The phrase "frameworks solve these problems" appeared 6 times across the 3,800-word post
(lines 220, 387, 443, 486, 616, 654).
Suggested consolidating into the conclusion section rather than repeating throughout.

**Grammar**
- Line 387: "ws" → "was"
- Line 28: "an form" → "a form"
- Line 614: "Problem 6 Regression Testing" → "Problem 6: Regression Testing" (missing colon for consistency)

The full skill implementation is available in my blog's GitHub repository: blog-editor/SKILL.md

Conclusion

Building this Claude skill has been like having a dedicated editor on call 24/7. While it won't replace the nuanced feedback of an experienced human editor, it provides a thorough editorial review that catches the majority of common issues.

If you're a technical blogger, I'd encourage you to try building a similar skill or feel free to copy and modify mine. The investment is minimal, it took one conversation with Claude to create, and the ongoing value is substantial. Every post I publish now gets that "second pair of eyes" review I've always wanted.

And in a pleasant bit of meta-circularity, this very post was edited by the skill it describes.