Specific prompting
Module 1+“Say who it’s for, what good looks like and what to avoid.”
Vague in, generic out. Specific asks are the single biggest lever on output quality.
Skills
You won't find lessons on these.
You'll pick them up because your project needs them, usually right after a reviewer points one out. Here's what that feedback sounds like, and roughly when it tends to show up.
Context management
Module 1+“Put that in CLAUDE.md so you don’t explain it every session.”
Claude starts each session fresh. What you write down, it remembers.
Verifying
Module 2+“How do you know it works? Show me you tested it.”
AI says “done” confidently. Checking is your job, and it is cheap.
Debugging with AI
Module 2+“Paste the error and ask Claude what it means before you try to fix it.”
Errors are information. Claude reads them faster than you can search for them.
Options, not answers
Module 3+“Ask for 20 names, not one. Then ask it to rank them against your positioning.”
The first answer is the average answer. Breadth, then judgment.
Adversarial AI
Module 3+“Have Claude play your most sceptical customer and tear this apart.”
AI agrees with you by default. Give it a role that argues back.
Giving references
Module 4+“Show it three sites you love and say what you love about each.”
Examples beat adjectives. “Modern and clean” means nothing; a screenshot means a lot.
Editing and taste
Module 5+“This reads like AI. Cut a third and add one thing only you would say.”
AI drafts. You decide what is good. That is the part nobody can automate.
Automation
Module 5+“You’ve done these steps three times. Turn them into a command.”
A pipeline you run by hand is a chore. A pipeline you save is leverage.
Safe autonomy
Module 6+“Let it draft the email. You press send.”
Give AI room to work and keep the irreversible steps for yourself.
Systems thinking
Module 6+“What triggers this, where is it recorded, who sees it?”
A company is a set of loops. Seeing them is how you know what to automate.