CLAUDE
Skills I have written for Claude, free to download. A skill is one SKILL.md file, sometimes with a few reference files beside it — instructions Claude loads when the task calls for them, and ignores when it does not.
Each one below has its own setup steps, because a skill for Cowork and a skill for code do not get installed the same way.
01cowork-setup
Set up Claude Cowork the right way.
A step-by-step walkthrough of how I set mine up, built from my own experience and a lot of ideas taken from people on the internet. It takes about twenty minutes. What you get back is a folder Claude reads at the start of every session, so you stop spending the first ten messages explaining who you are and what good looks like.
Your next steps
- 01Download the zip file. Keep it zipped, all five files go up together.
- 02Open Claude. Go to Settings → Customize → Skills → Add → Upload a skill.
- 03Create a folder on your computer called "Claude Cowork".
- 04Start a Claude Cowork session, point it to the "Claude Cowork" folder and prompt "help me set up Claude Cowork". The skill takes care of the rest.
Good luck!
- cowork-setup/SKILL.md
- cowork-setup/references/about-me-interview.md
- cowork-setup/references/anti-ai-writing-style.md
- cowork-setup/references/global-instructions.md
- cowork-setup/references/my-company-interview.md
02loop-engineering
Decide whether a task should be an agent loop at all, then design one that stops.
Runs a decision check before designing anything — can fresh feedback change the next step, can "done" be defined objectively, does the agent have a real way to verify. If any of the three fails it tells you to write a single sharp prompt instead. If they pass, it builds the loop around observe → choose → act → verify → record, with named end states so a hit budget cap is never dressed up as success.
A loop is only as good as its stop condition. The cheapest way for an agent to raise its own score is to move the target — so the goal has to live somewhere that changing it is visible.
- loop-engineering/SKILL.md
- loop-engineering/references/decision-checklist.md
- loop-engineering/references/playbook.md
Use them, change them, strip out the parts you disagree with. If one is wrong or you make it better, I would genuinely like to know.