I Built a Skill That Makes Claude Write Better Blogs, Here's How to Use It
You know that feeling when you ask Claude to write a blog post and it comes back sounding like a corporate press release from 2019? "In today's fast-paced world, it is important to leverage synergistic content strategies... blah blah blah"
Yeah. Not great.
The thing is, Claude is actually really good at writing, it just needs to be told how you want it to write. And that's exactly what Skills are for.
What Even Are Skills?
Think of Skills as a little instruction manual you hand to Claude before a task. Instead of re-explaining your preferences every single time ("be casual, no fluff, don't say 'in conclusion'"), you write it once as a Skill and Claude picks it up automatically whenever it's relevant.
Anthropic describes them as "folders of instructions, scripts, and resources that Claude loads dynamically to improve performance on specialized tasks." But honestly, a simpler way to think about it: Skills are reusable prompts that live in Claude's settings and activate automatically when needed.
Some Skills come pre-built from Anthropic, like enhanced Word, Excel, and PowerPoint creation. But you can also create your own custom ones for literally anything you do repeatedly.
Here's the Blog Writing Skill for Claude, Copy It
This is the blog writing skill I created that makes claude ai write better blogs, in human tone. Click here to download this "SKILL.md". Here’s the full content of the file:
---
name: blog-writer
description: "Write a complete blog post in a casual, conversational, human tone, like a smart friend explaining something clearly and engagingly. Use this skill whenever the user asks to write a blog post, article, or tutorial. Always output a downloadable Markdown file."
---
# Blog Writer Skill
Write engaging, human-sounding blog posts in the voice of a **smart friend explaining something**, not a textbook, not a corporate blog, not a generic AI article. Think: clear, warm, a little witty, no fluff.
## Voice & Tone Rules
- **Casual but not sloppy.** Use contractions, short sentences, conversational transitions ("Here's the thing...", "So what does that actually mean?", "Let's be real").
- **No AI clichés.** Never use: "In today's fast-paced world", "It's important to note", "In conclusion", "Dive deep", "Leverage", "Utilize", or hollow openers. Start with something that grabs.
- **Talk to the reader directly.** Use "you" and "i". Make them feel like you're having a conversation, not writing at them.
- **Explain like the reader is smart but new to this topic.** No over-explaining basic concepts.
- **Be concrete.** Use examples, creative analogies, short scenarios.
- **Allow personality.** Light humor, honest admissions ("I'll be upfront, this part is a bit dumb"), rhetorical questions, all fair game.
## Inputs
Gather these from the user's message (some may be optional):
| Field | Required | Notes |
|---|---|---|
| Topic | ✅ | The subject of the post |
| Target audience | Optional | Default: Mostly non-technical readers |
| Word count / length | Optional | Default: ask or infer from complexity |
| Key points to cover | Optional | If provided, make sure to hit them all |
If critical info is missing (e.g. no topic given), ask before writing. For optional fields, use your best judgment.
## Structure
Adapt structure to the post type, but a solid default for how-to / explainer blogs:
1. **Hook** — A question, bold claim, relatable scenario, or surprising fact.
2. **The Setup** — What problem does this solve? Why does it matter? Keep it tight (1 short paragraph).
3. **The Meat** — Main content. For tutorials: numbered steps with context. For explainers: logical flow with subheadings. Use `##` and `###` for sections.
4. **The Gotchas / Tips** — One section for caveats, common mistakes, or pro tips. Readers love this.
5. **Wrap-up** — Not a summary. A short punchy closing thought, encouragement, or forward-looking statement.
## Markdown Formatting
- Use `##` for main sections, `###` for sub-sections
- Use bullet lists for non-sequential items, numbered lists for steps
- Use **bold** to highlight key terms or important callouts (sparingly)
- Use `code blocks` for any code, commands, file paths, or technical strings
- Use blockquotes (`>`) for tips, warnings, or highlighted insights
- Keep paragraphs short — 2–4 sentences max
## Output Instructions
1. Write the complete blog post
2. Save it as a `.md` file in `/mnt/user-data/outputs/` with a slug-style filename based on the title (e.g. `how-to-write-better-hooks.md`)
3. Use `present_files` to deliver it to the user
4. After presenting, add a **one-line summary** of any assumptions made (length, audience, etc.) and offer to revise
## Quality Check (run mentally before saving)
- [ ] Does the opening sentence make you want to keep reading?
- [ ] Does it sound like a human wrote it?
- [ ] Are there any hollow filler phrases?
- [ ] Are the examples concrete and relatable?
- [ ] Is the length appropriate for the topic's complexity?
How to Set Up a Custom Skill (It's pretty simple)
No coding needed for a basic Skill. It's just a Markdown file SKILL.md with your instructions. Here's the exact process:
Step 1: Make sure Code Execution is on
Go to Settings → Capabilities and flip on "Code execution and file creation." Skills need this to function, it won't work without it.
Step 2: Create or Download this Skill markdown file
Click here to download or Create a markdown file called SKILL.md and paste the above instructions. That file is where all your instructions live.
Step 3: Upload it
Head to Customize → Skills, click the + button, then + Create skill, and choose "Upload a skill." Drop your SKILL.md file in.
Step 5: Toggle it on
Find your new Skill in the list and flip the switch. Done.
Now whenever you ask Claude to write a blog post, it'll automatically pick up the Skill and apply your instructions, no re-prompting required.
Worth Knowing
The description field matters a lot. Claude decides when to activate a Skill based on its description. If the description is vague, Claude might not know to use it. The one above is pretty specific ("whenever the user asks to write a blog post, article, or tutorial"), so it should trigger reliably.
that's it. thanx for reading :)