"I want to do X — which skills do I use?" This page answers that question for every scenario you'll face.
QUICK START
Find your goal below. Each card shows exactly which skills to use.
Start from zero: create repo, set up CI/CD, design system, and deploy — all in one session.
Find the real cause first, write a test to prove it, fix it, then deploy safely.
Analyze what to change, design with proven UX rules, preview the result, then build it.
Extract text, translate in batches, and auto-audit to catch errors before they go live.
Generate SEO articles, set up ad pixels (Meta, TikTok, Google), and measure what works.
Scan your codebase and auto-generate docs, SOPs, and API references.
STEP BY STEP
Click any scenario to see the exact steps. Follow them in order.
Start from nothing. In 1-2 chat sessions you'll have a live project with tests, CI, and auto-deploy.
Don't let AI guess randomly. This workflow finds the real cause first, proving it with a test before fixing anything.
One wrong translation key = broken page. This skill extracts, translates, and double-checks everything automatically.
Need blog articles + ad tracking? This chain auto-writes, quality-checks, deploys, and sets up tracking pixels.
Don't jump straight to code. Analyze what to change, design it properly, preview it visually, then plan the build.
⚠️ TROUBLESHOOTING
Hit a wall? Here's the fix for every common problem.
Long chats fill up the AI's memory. It "forgets" earlier instructions and starts going off track.
You're halfway done and realize the approach won't work, or requirements changed.
The AI guesses randomly instead of understanding why it's broken. Each attempt makes it worse.
Multiple GitHub/Cloudflare accounts? One wrong push = your code ends up in someone else's project.
Large codebase (200+ files). AI reads a few files but misses the rest — changes break things it never saw.
AI puts passwords or API keys directly in code. Push to a public repo = instant security breach.
🏆 GOLDEN RULES
Follow these and your AI-assisted development will be predictable, safe, and fast.
Identity First.Verify account (cm-identity-guard) before push/deploy. Every. Single. Time.
Design Before Code. Always have an approved plan before typing. Use cm-planning.
i18n Day 0. Always consider multi-language from the brainstorm step, not as an afterthought.
Test Before Code. RED → GREEN → REFACTOR. No exceptions. cm-tdd enforces this.
Evidence Over Claims. Only trust terminal/test output, not AI "saying" it's done.
Deploy via Gates. 8 gates must pass sequentially. Any fail = STOP. cm-safe-deploy.
Safe Secrets. Never commit secrets. Pre-commit hooks protect every push. cm-secret-shield.
Parallel Power. Use parallel execution for i18n or multi-bug fixes. cm-execution.
Working Memory. Read CONTINUITY.md at session start, update at end. cm-continuity.
Install in 30 seconds. Follow the playbook. Ship with confidence.
Have questions? Check the FAQ →