Getting Started
Getting Started with Claude Code & ClaudeKit Engineer Kit
Claude Code is an agentic coding environment from Anthropic. ClaudeKit Engineer Kit transforms Claude from assistant into a true development partner with skills, hooks, and workflows.
What is Claude Code?
- Agentic coding environment: reads files, runs commands, auto-solves problems.
- Key constraint: context window fills up → performance degrades. Use /clear between unrelated tasks.
- Claude Code works best when you understand how it manages context.
What is ClaudeKit Engineer Kit?
- A .claude/ folder containing workflow templates: Skills, Hooks, Agents, Rules.
- Philosophy: YAGNI + KISS + DRY — simplify work, avoid unnecessary complexity.
- Engineer Kit transforms Claude from assistant into a true development partner.
Memory Loading Order
Claude reads and concatenates files in order, from global → project → path-specific:
~/.claude/CLAUDE.md Global — all projects ./CLAUDE.md + CLAUDE.local.md Project root .claude/rules/*.md Topic rules (lazy-load if path-scoped) Tip: Keep CLAUDE.md < 200 lines. Split domain rules (api, testing, security) into .claude/rules/ to avoid context bloat.
CLAUDE.md vs rules/ — What's the difference?
| CLAUDE.md | .claude/rules/*.md |
|---|---|
| General instructions for whole project | Modular instructions, split by topic/path |
| Always loads into context at launch | No paths: load at launch · Has paths: lazy-load when files match |
| Target: under 200 lines | One file per topic (code-style.md, testing.md...) |
| Build commands, general conventions | Domain rules: API, frontend, security, DB... |
| Single file only | Multiple files, recursive subdirectories |
Use YAML frontmatter with paths field so rules only activate when Claude touches matching files:
Rules without paths field load unconditionally. With paths, triggers when Claude reads files matching pattern.
Best Practices for Creating Rules
DO THIS
- Split rules by domain/concern (i18n-rules.md, testing-rules.md...)
- Write concisely, bullet/checklist — Claude parses fast
- Kebab-case, descriptive names (api-error-handling.md)
- Include examples (good vs bad)
- Reference exact paths when applying to specific folders
AVOID THIS
- Writing verbose rules like documentation
- Duplicate content between CLAUDE.md and rules/
- Creating rules for YAGNI things (1-time use)
7 core commands for beginners
/ck:brainstorm Explore ideas and solutions before coding /ck:plan Create detailed implementation plans with phases /ck:cook Execute plan — build features step by step /ck:fix Debug and fix bugs systematically /ck:ask Quick Q&A about code or architecture /ck:test Run tests and analyze coverage /ck:watzup See summary of what was done in session Do This
- Give Claude clear verification criteria so it can self-check results.
- Follow the flow: Brainstorm → Plan → Cook — don't jump straight to code.
- Provide specific context: file paths, error messages, expected behavior.
- Clear context frequently between unrelated tasks.
Avoid This
- Kitchen-sink sessions: cramming too many tasks into one session.
- Repeated corrections without clearing context → Claude repeats mistakes.
- Skipping planning, jumping straight into implementation.
- Trust without verify — always check Claude's output.
- Vague prompts: "fix this" instead of describing the specific problem.
Real-World Use Cases
Add user settings page with dark mode toggle
Form validation not showing error
Join new team, understand codebase fast
→ 30min instead of 2 days
Old 500 LOC module needs cleanup
Read more details
Read the full article for workflow patterns, weekly skills roadmap, and cheat sheet →