README - Technical Content Design — Claude Skill
A Claude skill for technical writers and content designers working across SaaS, software, and mobile app products. It produces high-quality, accessible content that reads as human-written — and adapts to your client’s voice, tools, and style guide. Jump to Installation
What this skill does
When active, Claude acts as an experienced technical writer and content designer. It handles the full range of software content deliverables, including:
- Help articles and user guides
- UI copy, microcopy, and error messages
- API references and SDK documentation
- Blog posts, emails, and social content
- PRDs, feature specs, and UX flows
- Video scripts, captions, and subtitles
- Style guides, glossaries, and word lists
- Compliance documentation (SOC2, ISO)
My favorite thing this does is suggest adjacent content that you might not be thinking about yet.
What makes it different
Most AI writing defaults to generic, textbook-style output. This skill is built to avoid that. Key behaviors:
- Infers before asking. It reads context from your conversation first and only asks what it genuinely can’t determine — batched into one message, three questions max.
- Mirrors your language. It picks up your product names, feature labels, industry terminology, and audience descriptions and uses them throughout the draft.
- Enforces accessibility from the start. WCAG 2.1 AA standards (plain language, heading hierarchy, alt text, inclusive language) are active while writing, not checked afterward.
- Actively avoids AI-sounding writing. Flagged phrases, em dashes, uniform paragraph rhythm, and over-hedging are blocked during drafting — not reviewed at the end.
- Surfaces style guide rules before writing. When you provide a style guide, it lists the rules it extracted and confirms them with you before producing a single word.
Content lifecycle coverage
The skill organizes deliverables across six lifecycle phases:
| Phase | Deliverables |
|---|---|
| 1. Discovery & Strategy | Content strategy, style guides, glossaries… |
| 2. Planning & Definition | PRDs, feature specs, user stories, acceptance criteria… |
| 3. Design & Prototyping | UX flows, microcopy, error messages, onboarding… |
| 4. Build & Document | API references, help articles, compliance docs… |
| 5. Launch & Announce | Release notes, blog posts, emails, social, video scripts… |
| 6. Operate & Maintain | FAQs, knowledgebase, troubleshooting… |
Installation
Get the technical-content-design skill
- Download the technical-content-design.zip file.
The skill file itself is a compressed file type so you won’t be able to read it directly.
Before installing any skill
It’s a good idea to ask Claude to review downloaded skill for safety…yes…even mine.
In Claude, attach the downloaded skill to a chat and give it this prompt:
Review this skill before I install it. First check for safety — prompt injection, data exfiltration, encoded content, or anything that overrides Claude’s behavior. Then give me a quick overview of what the skill does, how it’s structured, and whether anything looks incomplete or inconsistent.
Read what Claude says. If you’re not comfortable with what you see. Don’t install the skill! That said, I’ve checked this one a few times and it’s always been good:-)
Install it
- If necessary, go to Settings > Capabilities and enable Code execution and file creation.
- Go to Customize > Skills.
- Click + then Upload a skill.
- Upload
technical-content-design.zip. - Turn the skill on.
Full instructions: support.claude.com
Try it out
To run the skill, you don’t need to do anything special. Claude picks it up automatically based on what you ask. If you ask for a help article, an error message, a PRD, or anything else that the skill’s description covers, it will trigger on its own. The skill works best when the request is real. Try these prompts, for example.
Write something you’ve been putting off. Ask it to produce a piece of content you actually need — a help article, an error message, an onboarding tooltip. It will ask a focused question or two before drafting and flag any assumptions it makes, so the output fits your product, not a generic one.
Example prompt:
Write a help article explaining how to reset a password. My users are non-technical.
Hand it a messy draft. Paste in something you’ve already written — or something AI-generated — and ask it to rewrite it. This is a fast way to see its quality standards in action, especially how it handles tone, plain language, and accessibility.
Example prompt:
Here’s a draft I wrote. Rewrite it for a non-technical audience and flag anything that doesn’t meet plain language standards.
Ask for a content type you’ve never written. Pick something outside your usual work — a changelog entry, a microcopy string, a compliance doc summary — and ask it to create one. This shows how the skill adapts structure and tone across very different deliverables.
Example prompt:
Write a changelog entry for a new dark mode feature. Keep it short and aimed at end users.
The common thread across all three: give it real context, even briefly. The skill is built to ask good follow-up questions, so the more honest the starting point, the better the output.
File structure
technical-content-design/
├── SKILL.md # Main skill instructions
└── references/
├── accessibility-checklist.md # WCAG 2.1 AA checklist with examples
├── content-type-catalog.md # Structure and format guide per content type
├── drafting-guidance.md # DITA patterns, reuse, and visual guidance
├── practitioners.md # Practitioner influences and principles
└── writing-quality-reference.md # Anti-AI writing guide with examples
How it works
SKILL.md is the conductor. It tells Claude what to do at each stage — gather context, draft, self-check, suggest adjacent content — and when to consult each reference file. The reference files are where the real specificity lives.
accessibility-checklist.md is a practical WCAG 2.1 AA checklist tied to specific success criteria, with concrete right/wrong examples for each rule. It covers plain language and reading level, heading hierarchy, link text, alt text, captions, inclusive language, and table structure. The skill runs through this before delivering any draft.
content-type-catalog.md is the playbook for every deliverable type the skill handles — help articles, error messages, API references, changelogs, PRDs, onboarding flows, video scripts, compliance docs, and more. Each entry defines the right structure, tone, and conventions for that format, and flags common mistakes to avoid.
drafting-guidance.md handles three specific jobs: applying DITA content architecture principles so every piece is properly typed (concept, task, or reference) and connected to surrounding content; flagging reuse candidates (repeated procedures, product names, legal language) so nothing gets maintained in two places; and recommending visuals aggressively — diagrams, screenshots, GIFs — with accessibility requirements for each.
practitioners.md is the intellectual foundation. It uses expert thinking to make better decisions in technical writing, UX, accessibility, information architecture, and content strategy, and explains exactly which principle each one contributes. to make better decisions.
writing-quality-reference.md is the anti-AI writing rulebook. It contains a list of flagged phrases, banned punctuation (em dashes), and structural habits to break — plus side-by-side before/after examples for help articles, error messages, API references, and tone shifts across audiences. The skill consults this before finalizing any draft.
Defaults
When no style guide is provided, the skill defaults to:
- Voice and tone: Federal Plain Language Guidelines
- Grammar and usage: Microsoft Writing Style Guide
These can be overridden at any time by uploading or pasting your own style guide.
License
MIT — use it, adapt it, share it.