She catches
what you
miss
You prompt. Claude codes.
Claudia makes sure it's actually good.
Whether you're vibe coding your first app or shipping to production.
npm i -g claudia-mentor How she works
Three ways
she shows up
You don't change anything about how you work. Claudia runs alongside Claude Code and steps in when it matters.
Automatic checks
Every time Claude writes a file, Claudia scans it. Leaked API keys get blocked. Bad patterns get flagged. You see what's wrong and how to fix it.
She joins your conversation
Say "I'll use MongoDB for this" and Claudia jumps in: "Actually, Postgres handles this better — here's why." Claude sees her advice and adjusts.
Ask her directly
11 commands — /claudia:ask, explain, review, why, health, wtf, where, resume, setup, start, shortcuts. She gives you the answer, the tradeoffs, and the reasoning.
Automatic protection
You didn't know this
was dangerous. She did.
She's watching
every file you write.
15 hooks. Always on.
You don't have to
remember any of this
8 hooks guard every file Claude writes. 7 more watch your session and teach you as you go. You never run them — they just work.
Code guardrails
Proactive coaching
Ask her anything
You don't need the answer.
Just the question.
Better prompts = better code
The difference
between "fix the
auth" and getting
what you want
Most vibe coders leave 80% of Claude's power on the table because they don't know how to ask. Claudia notices when your prompt is too vague and shows you how to rewrite it.
What you typed
"Fix the auth"
What Claudia suggests
"Refactor the auth middleware to use async/await, add error handling for expired tokens, and return 401 responses with JSON error codes."
Never coded before?
She meets you
where you are
Set experience to "beginner" and Claudia transforms. No command lists. No jargon. She waits until you need her, then shows up with exactly the right thing.
{ "experience": "beginner" }Stuck detection
Type "I'm stuck" or "idk" and she asks one clarifying question, then gives you one small next step. No overwhelm.
Run suggestions
Claude creates a file, Claudia tells you how to run it. open index.html, python3 app.py — the stuff you wouldn't know to type.
Milestones
"First commit. Your code has a save point now." She notices when you hit real milestones and tells you what they mean.
Progressive reveal
Commands appear only when they're relevant. Write a file, she mentions /claudia:explain. Hit an error, she mentions /claudia:wtf. No firehose.
Next-step suggestions
After Claude finishes something, Claudia suggests 2-3 things you could do next. No dead ends.
Already know what you're doing?
She's not just
for beginners
Claudia started as a teaching tool. But 15 hooks watching every file write turns out to be useful for everyone.
15 hooks, zero config
Secret detection, dependency audits, Docker linting, CSS checks, license scanning — all running automatically on every file write.
Blocks before it lands
Hardcoded AWS keys and .env writes get stopped before they hit your repo. Not warned — blocked.
10 knowledge domains
Security, databases, infrastructure, APIs, frontend, testing, performance, DevOps, data modeling, and prompt coaching. All opinionated.
Open source, tested
MIT licensed. 234+ tests. No telemetry, no accounts, no network calls. Just a Claude Code plugin that does its job.
Ship with
confidence
One install. She handles the rest.
Free, open source, MIT licensed.