/think
Think through a problem on paper before acting
$ npx -y skills add markmdev/meridian --agent claude-codeHow it fires
How this command gets triggered: by you, by Claude, or both.
- Fires itselfClaude auto-loads it when your prompt matches the work.
- You can call itInvoke it directly when you want it.
- Slash command
/think
Context preview
What this command does when you run it.
Think through a problem on paper before acting
Command definition
think.mddescription: Think through a problem on paper before acting argument-hint: [LINES] [TOPIC]
Think Mode
You have been asked to think through a problem before acting.
**Arguments:** $ARGUMENTS
Parse the arguments:
- First number found = line limit (default: 50 if not specified)
- Remaining text = topic to think about (optional)
**Instructions:**
1. Create a scratch file: `.meridian/.scratch/thinking-YYYY-MM-DD-HH-MM-topic.md`
- Use current timestamp
- Slugify the topic for the filename (or use "exploration" if no topic)
2. Write your thoughts to this file. You have **{line_limit} lines** to think.
- Stream of consciousness — explore angles, question assumptions
- What could go wrong? What don't you know? What are you assuming?
- Consider multiple approaches, note tradeoffs
- Messy is fine. Contradictions are fine. Dead ends are fine.
- This is exploration, not planning.
3. End with a brief "## Synthesis" section summarizing:
- Key insights
- Open questions
- Recommended next step
4. After writing the thinking file, tell the user what you figured out and ask how to proceed.
**The point:** Thinking on paper catches issues early and produces better solutions than jumping straight to action. Use the line limit — don't stop early, don't exceed it significantly.
Meridian makes Claude Code more reliable on real projects. It adds persistent project context, smarter session handoff, and lightweight workflow enforcement so Claude is less likely to lose the plot halfway through a long task.