A collection of structured prompt guides for Claude Code that turn the CLI into a consistent, context-aware coding assistant.
Each guide is fetched on-demand via a keyword trigger in your ~/.claude/CLAUDE.md.
| Keyword | Guide |
|---|---|
debugging |
Debug a problem step by step |
review |
Review code for quality, security & performance |
planning |
Plan a feature before writing any code |
explain |
Get a concept explained at your level |
refactor |
Restructure code without changing behavior |
testing |
Write meaningful tests for your code |
Add trigger lines to ~/.claude/CLAUDE.md:
Wenn ich "debugging" sage, fetche und befolge: https://raw.githubusercontent.com/RobbyRunge/claude_code_helper/main/debugging_guide.md
Wenn ich "review" sage, fetche und befolge: https://raw.githubusercontent.com/RobbyRunge/claude_code_helper/main/code_review_guide.md
Wenn ich "planning" sage, fetche und befolge: https://raw.githubusercontent.com/RobbyRunge/claude_code_helper/main/feature_planning_guide.md
Wenn ich "explain" sage, fetche und befolge: https://raw.githubusercontent.com/RobbyRunge/claude_code_helper/main/explain_guide.md
Wenn ich "refactor" sage, fetche und befolge: https://raw.githubusercontent.com/RobbyRunge/claude_code_helper/main/refactor_guide.md
Wenn ich "testing" sage, fetche und befolge: https://raw.githubusercontent.com/RobbyRunge/claude_code_helper/main/testing_guide.md
Claude will fetch the matching guide whenever you use the keyword in a conversation.
A step-by-step debugging workflow that adapts to your goal:
- Clarifies project type (learning / professional / exam)
- Selects stack (Angular, Django, React, other)
- Chooses mode: J (guided learning) or N (direct solution)
- Always explains the root cause, never just fixes the symptom
- Ends with a link to the official documentation
A structured code review that covers all the areas that matter:
| Area | Focus |
|---|---|
| S Security | Injection, auth gaps, data leaks |
| P Performance | N+1 queries, unnecessary re-renders |
| Q Quality | Naming, structure, readability |
| B Best Practices | Stack conventions, typing, error handling |
Supports the same two modes as the debugging guide — learn-focused or straight to the fix.
A structured planning session before writing any code:
- Covers backend (models, migrations, endpoints, serializers, signals) and frontend (components, state, routing, error handling)
- Explicitly walks through edge cases and risks
- Delivers a prioritized checklist with recommended implementation order
- Supports learn mode (think it through together) and direct mode (give me the plan)
Concept explanations tailored to your current knowledge level:
- Asks what you already know before explaining — no re-explaining the obvious
- Three depth levels: O (overview), D (internals), P (practical example in your project context)
- Always includes an analogy and a comprehension check at the end
- Distinguishes when to use one concept vs. another
Structured refactoring that improves code structure without changing behavior:
- Identifies candidates: readability, duplication, responsibility, complexity, framework anti-patterns
- Names the refactoring technique used (Extract Function, Rename, etc.)
- Always includes a behavior-check at the end to confirm nothing broke
- Supports learn mode (explain each step) and direct mode (deliver the refactored code)
Helps write tests that give real confidence, not just coverage numbers:
- Recommends the right test type (unit / integration / E2E) for the code at hand
- Follows Arrange / Act / Assert structure throughout
- Explicitly calls out what NOT to test (framework behavior, trivial getters)
- Includes a coverage-check at the end: what's missing and what to tackle next
git_guide.md— commit messages, branch naming, PR descriptions, when to squashsecurity_guide.md— OWASP checklist, Django security pitfalls, JWT edge casesperformance_guide.md— profiling-first approach, DB query optimization, frontend render performance
All guides currently support:
- Angular / TypeScript
- Django / Python
- React / TypeScript