Get coached by AI
on how you are using AI.
A free Claude Code skill that scores your recent prompts on Opal University's CLEAR framework and tells you the one move that would level you up. One file. Zero install friction. No account needed.
Why this exists
Everyone is learning AI alone. Nobody knows what good looks like.
Your boss told you to figure out AI. Every LinkedIn post makes you feel 10 steps behind. There is no playbook. No peer at your company is ahead of you to copy.
The thing nobody tells you is that prompting is a learnable skill. There are patterns that make AI respond better, and patterns that waste your time. You just have not had anyone coach you on the difference.
This skill is that coach. In one minute, you will know where you are and what to try next.
How it works
Three steps. Under a minute.
Install the skill
One line in your terminal. The skill is a single markdown file. No account, no token, no backend.
Prompt Claude Code like you always do
Keep working. The skill stays quiet until you ask for it. Your prompts never leave your machine.
Say "coach me"
Claude pulls your recent prompts, scores them on CLEAR, tells you the one move that would level you up.
What you will see
A report card, not a grade.
When you say "coach me," Claude writes you something like this. Specific praise. One real opportunity. One move to try next.
Your AI Readiness Report
Level: Operator
What is working
Your second prompt named the file (lib/auth.ts) and quoted the error. That is pro-level context and it is why Claude got it right on the first try.
Biggest opportunity
Your Actionsscore is the weakest. You asked Claude to "make it better" twice without saying what better means. Try this rewrite of your last prompt:
Try this on your next prompt
Add one line up front that says what format or length you want back. That alone usually bumps 2 points.
The 5 dimensions
CLEAR
- C
Context
Did you name the file, the error, the constraint?
- L
Layout & Logic
Is the ask organized, or a wall of text?
- E
Examples
Did you show what good looks like?
- A
Actions
Did you say what format, length, scope, tone?
- R
Review & Refine
Constraints, edge cases, "do not do X"?
The 5 levels
Curious → Principal
Curious
Just getting started. Every prompt is practice.
Practitioner
Finding your rhythm. Prompts are landing more often.
Operator
Steady and reliable. You know what to ask for.
Coach
You could teach this. First-try results are the norm.
Principal
Top 1%. You treat prompting like engineering.
Install
One line. Paste it.
This creates ~/.claude/skills/opal-coach/SKILL.md and downloads the skill. Restart your Claude Code session after running it.
mkdir -p ~/.claude/skills/opal-coach && curl -fsSL https://university.optimizely.com/skills/opal-coach/SKILL.md -o ~/.claude/skills/opal-coach/SKILL.mdAfter you paste it
Open a new Claude Code session, work on anything for 3 to 5 prompts, then say "coach me." That is the whole thing.
Want to see the file first?
Read the SKILL.md. It is plain English. No tricks, no telemetry, no network calls.
FAQ
Real questions, answered.
Your next move
Install the skill.
Find out where you are.
One minute from now, you will have a report card. Tomorrow, you will prompt better. Next week, your team will notice.