name: thoughtful-questions
description: |
Applies when you need to ask the user clarifying questions about a task or requirement.
Only ask meaningful questions that require human judgment or preference — never ask things
you can validate, check, or figure out yourself. When you have multiple questions, use
the execute_command tool to self-invoke /answer for a structured Q&A experience.

Thoughtful Questions

When you need clarification from the user, be deliberate about what and how you ask.

Rules

1. Only Ask What You Can't Answer Yourself

Before asking, consider: Can I figure this out myself?

• Can I check the codebase for conventions? → Do it
• Can I look at existing files for patterns? → Do it
• Can I try something and see if it works? → Do it
• Can I make a reasonable default choice? → Do it

2. Ask Meaningful Questions

Good questions require human judgment, preference, or domain knowledge:

✅ Meaningful:
- "Should this be a breaking change or should we maintain backwards compatibility?"
- "Do you want this optimized for speed or readability?"
- "What's the business logic when X happens?"

❌ Obvious/wasteful:
- "Do you want me to handle errors?" (yes, obviously)
- "Should I add comments?" (use judgment)
- "Does this file exist?" (check yourself)

3. Multiple Questions? Use /answer

When you have multiple questions to ask, don't make the user answer inline. Instead:

1. Format questions clearly — End each with ? so the extractor can find them
2. Self-invoke /answer — Use the execute_command tool to run /answer after your response

Example flow:

You: "I have a few questions about the implementation:

1. Should we use REST or GraphQL for the API?
2. Do you want authentication on all endpoints or just write operations?
3. What's the expected request volume — do we need rate limiting?"

[execute_command tool: command="/answer", reason="Opening Q&A interface for the user to answer these questions"]

The /answer tool extracts your questions and presents an interactive UI where the user can navigate and answer each one efficiently.

Philosophy

The user's time is valuable. Every question you ask should be meaningful.

• Explore first — read code, check files, try things
• Decide what you can — use good defaults and conventions
• Ask only what matters — things that genuinely need human input
• Make it easy to answer — use /answer for multiple questions