قائمة التحقق اليومية لمستخدم Claude المحترف: ٣٠ عادة للإنتاجيةDaily Claude Power User Checklist: 30 Productivity Habits

Key terms

CLAUDE.md

A text file at the root of your project that Claude Code reads automatically at the start of every session. It is the project's persistent memory: your decisions, preferences and rules.

/clear

A Claude Code command that wipes the current conversation context so the next task starts clean.

claude -p

Runs Claude Code without the interactive interface. You send it a request from the terminal and it prints the answer.

Prompt library

A personal folder of your best tested prompts saved as text files, organized by task type.

The steps

1. 1

   Step 1

   Create your project CLAUDE.md

   CLAUDE.md is the most powerful tool in Claude Code. Claude reads it automatically at every session start and uses it as persistent context.

   Create one per project and keep it updated. You should never have to explain the same thing twice.

   Open the terminal at your project root, run the commands below, then edit the content to match your project.

   PromptTerminal commands

   Copy

   cd ~/projects/my-app
   cat > CLAUDE.md << 'EOF'
   # Project: my-app
   ## Stack
   Node.js 20, TypeScript strict, PostgreSQL, Tailwind CSS
   ## Code Conventions
   - Functional patterns preferred over classes
   - All async functions must handle errors explicitly
   - Arabic comments for business logic, English for technical
   ## Output Preferences
   - Always show diffs, not full file rewrites
   - Explain reasoning before writing code
   - When uncertain, ask one clarifying question
   EOF
   cat CLAUDE.md

   Tip: CLAUDE.md works far better with concrete examples than with abstract rules alone.

2. 2

   Step 2

   Set up the session before you type

   A great session starts before your first prompt. Two minutes of setup save an hour of back and forth.

   The six morning habits: review your CLAUDE.md before opening any conversation. Define your session goal in one sentence. Launch Claude Code from inside your project directory. Use /clear between unrelated tasks. State the output format you want in the first message. Avoid running several sessions on the same project at once.

   Paste the primer below at the start of every work session so Claude is oriented instantly.

   PromptDaily context primer

   Copy

   You are a senior [ROLE: for example backend engineer, product manager, technical writer] helping me work on [PROJECT NAME].
   Today is [DATE]. My single goal for this session is: [ONE SPECIFIC SENTENCE].
   
   Project context:
   Stack: [YOUR TECH STACK]
   Current phase: [for example building the auth module, fixing a production bug, writing API docs]
   Last session we: [ONE OR TWO SENTENCE SUMMARY]
   Decisions already made that are not open for debate: [LOCKED DECISIONS]
   
   My preferences for this session:
   Output format: [for example diffs only, bullet points, full code blocks with explanation]
   Language: [LANGUAGE FOR EXPLANATIONS], English for code and terminal commands.
   When you are unsure about something, ask me exactly one clarifying question before proceeding. Do not guess.
   Do not rewrite entire files when a targeted diff will do.
   
   Acknowledge this context in one sentence and suggest the most logical first action to take right now.

3. 3

   Step 3

   Write prompts with discipline

   A bad prompt is not Claude's problem. Prompt writing is a technical skill that takes practice.

   The strongest pattern is role, then task, then format. Say who Claude is, then the task, then the shape of the answer.

   The six prompting habits: use role, task, format in every prompt. State the expected length. Say explicitly what you do not want. Give at least one example when you need high precision. Iterate inside the same conversation instead of restarting. Use XML tags to structure complex inputs with several parts.

   Define what a correct answer looks like before you ask for the solution. Claude fills empty space with assumptions, and explicit constraints steer it.

   PromptRole, task, format prompt

   Copy

   You are a [ROLE].
   Task: [EXACTLY WHAT YOU WANT].
   Format: [TABLE, BULLETS, CODE ONLY], no longer than [LENGTH].
   I do not want: [WHAT TO AVOID].
   Example of the output I want:
   <example>
   [ONE EXAMPLE]
   </example>

4. 4

   Step 4

   Build your personal prompt library

   Most wasted time comes from rewriting the same prompt from memory. Power users save tested prompts as text files.

   The simplest structure: one folder per task type, such as code writing, code review, docs and analysis. Each file has a descriptive name and one current version ready to use. The commands below are for macOS.

   The six organization habits: save the best prompt you used each day before you sleep. Organize prompts by task type, not by date. Keep CLAUDE.md in the same git repository as the project. Name prompt files so their purpose is clear at a glance. Review the library once a week and delete what keeps failing. Add a short note to prompts that worked exceptionally well, saying why.

   PromptPrompt library commands

   Copy

   mkdir -p ~/.claude-prompts/{code,review,docs,analysis,daily}
   cat ~/.claude-prompts/daily/session-start.md | pbcopy
   find ~/.claude-prompts -name '*.md' | sort

   Tip: Keeping CLAUDE.md in git preserves how your understanding of the project evolved.

5. 5

   Step 5

   Give the right context at the right moment

   Context is the main currency with Claude. The sharper it is, the better the output. Too much unstructured context hurts.

   The six context habits: share full error messages, never summaries. Name the file and line number when discussing specific code. Attach recent commits when asking about how the project evolved. Persist technical decisions in CLAUDE.md. Update CLAUDE.md at the end of every session with any important decision. Use /clear when switching to a completely different project or topic.

   The commands below pipe your git history straight into Claude with claude -p for fast, context rich answers.

   PromptGit log as live context

   Copy

   git log --oneline -10 | claude -p "Based on these commits, summarize what changed in the last few days and flag any patterns or risks you notice"
   
   git diff HEAD~1 HEAD | claude -p "Review this diff for bugs, missing error handling, and style inconsistencies. Be specific about file names and line numbers."
   
   git log --oneline --since='yesterday' | claude -p "Write a two sentence standup update based on these commits. Professional tone, past tense, no jargon."

6. 6

   Step 6

   Review your day in five minutes

   Five minutes at the end of each day is the difference between a user who improves and one who repeats the same mistakes.

   Ask Claude for a summary of what you accomplished before closing the conversation. That summary becomes the opening context of your next session.

   The six end of day habits: ask for a summary of what you accomplished. Save today's best prompt to your library. Note what failed and why in two sentences at most. Update CLAUDE.md with new rules and decisions. Plan tomorrow's first task and write it down now. Commit any prompt files or CLAUDE.md you changed today.

   PromptEnd of day debrief

   Copy

   Before we close this session, give me a structured end of day debrief using this exact format:
   
   <accomplished>
   Every concrete thing we completed today, with file names, function names or feature names where relevant. Bullet points. Be specific.
   </accomplished>
   <in_progress>
   Everything we started but did not finish, with the exact state of each item so I can pick it up tomorrow without losing context.
   </in_progress>
   <blockers>
   Any technical or conceptual problems still unresolved, what we tried and exactly why it did not work.
   </blockers>
   <learnings>
   Anything new you learned about this codebase, my working preferences or the problem domain that I should add to CLAUDE.md.
   </learnings>
   <tomorrow>
   One specific first task for tomorrow that I can start within thirty seconds without any research.
   </tomorrow>
   
   Format the entire output as a clean markdown document I can save directly to my project journal.

   Tip: When Claude repeats a mistake, add the fix as a rule in CLAUDE.md right away, not just in the prompt.

7. 7

   Step 7

   Start with two habits this week

   Thirty habits sounds like a lot, but each one saves more time than it costs.

   Do not try to change everything at once. Incremental adoption wins. Start with the morning setup and the end of day review, and the rest follows.

   Block a fixed slot each week to review your prompt library: delete the failures and improve the winners.

Ready prompts to copy

You are a senior TypeScript engineer. Refactor the function below to be more readable and efficient. Keep the same external behavior. Show only the diff. Explain each change in one bullet point after the diff.

[PASTE FUNCTION HERE]