I fired my team and hired Claude Opus 4.6
The complete guide to Skills, Cowork, and Plugins.
In this issue:
→ What Skills are and how to build your first one
→ 10 prompts you can copy and use in Cowork today
→ 6 workflows for content, sales, legal, and visualisation
→ 5 Skill libraries with 1000s of free templates
Anthropic’s Cowork launch wiped $285 billion off global markets in a single day.
Investors coined a new term: “SaaSpocalypse.”
Not because of a new chatbot. Because Cowork turns Claude from a chat window into something that reads your files, organises your folders, drafts your documents, and runs multi-step workflows you used to delegate to a team.
Today we’re fixing that. Sabahudin and I are breaking down Skills and Cowork together. No technical background needed.
If you’ve ever written a checklist, you have everything you need.
Claude Opus 4.6
Quick note before we start.
Claude Opus 4.6 launched this week. The context window jumped to 1M tokens.
That means Claude can now: → Read an entire book in one go → Process your whole codebase at once → Hold conversations that don’t cut off halfway through
For months, Claude’s small context window was its biggest weakness. Conversations got cut short. Documents got lost. Every other frontier model handled more.
That’s gone now.
What still lets it down:
The API is still expensive
Gemini handles images and video better
But for strategic thinking, long-form writing, deep research, and coding? Nothing touches it right now.
For everything in this newsletter, Opus 4.6 is the model to use. Select it in the model dropdown (Pro/Max only).
Skills require a Pro, Max, Team, or Enterprise plan. Team and Enterprise users: your admin must enable Skills first in organisation settings.
Set Up Claude Properly First
Before building Skills, turn on these three settings. They make everything work better.
1. Memory (Settings > Personalisation > Enable Memory)
Not on by default. Once enabled, Claude remembers preferences across all chats. Add custom instructions in “User Preferences.” Works with Projects too.
2. Extended Thinking (Enable in settings or per-message)
Claude reasons step-by-step before answering. You see the thinking in grey text. Uses more tokens but way better accuracy.
3. Projects (For long-running work)
Organise work by client, research topic, or content series. Upload files as persistent context. Add Project Instructions (like Skills but project-specific). All chats in the project share context.
Memory works across Projects. Your preferences follow you everywhere.
What a Skill Is (and Isn’t)
Think about training a new employee.
Day 1, you explain how to write your reports.
Day 2, you explain it again.
Day 3, same thing.
By week two, you’re exhausted.
Now imagine you wrote everything down once. Your new hire reads it, follows it, delivers the work exactly how you want. Every time.
A Claude Skill is that written guide.
You create it once. Claude follows it from that point forward.
The difference between a Skill and a prompt: a prompt disappears when the conversation ends. A Skill persists. It lives in your account, ready to activate whenever the task matches.
Skills also load smarter. Claude reads the name and description first, then pulls in the full instructions only when needed. Your context window stays clean. Your tokens go further.
Three things Skills do:
Package your workflows. Monthly reports, client emails, content briefs. Write the process once, use it forever.
Add specialised knowledge. Your templates, your tone, your rules. The difference between a generic assistant and one that sounds like your team.
Stack with built-in abilities. Your lead qualification Skill works alongside Claude’s Excel Skill. One prompt scores 20 prospects, colour-codes by priority, and exports a spreadsheet. Two Skills. One prompt. Output ready to send.
Skills follow an open standard.
This means you can build in the web app. Use in Claude Code or Cowork. Share across your team. You’re building an asset, not a prompt that dies when the chat ends.
How to Build Your First Skill
No code required.
