clarify-first
Ask clarifying questions and get approval before implementation.
When & Why to Use This Skill
The 'clarify_first' Claude skill is a vital behavioral framework designed to enhance AI reliability by implementing a human-in-the-loop validation process. It ensures that the AI agent identifies high-risk or ambiguous tasks and seeks explicit user approval before proceeding, effectively preventing costly errors while maintaining momentum on low-risk, straightforward implementations.
Use Cases
- Complex Code Refactoring: The agent identifies potential side effects in a legacy codebase and asks for architectural approval before applying changes.
- Ambiguous Data Transformation: When faced with conflicting data requirements, the agent pauses to clarify the intended output format instead of making assumptions.
- High-Stakes Automation: Before executing irreversible actions like deleting cloud resources or sending bulk emails, the agent presents a summary for final user sign-off.
- Requirement Gathering: During the initial phase of a project, the agent suggests improvements to a prompt and asks for confirmation to ensure the final build aligns with user expectations.
| name | clarify_first |
|---|---|
| description | Ask clarifying questions and get approval before implementation. |
| short-description | Clarify before build |
Purpose
Clarify only when needed without blocking progress.
Steps
- Ask clarifying questions only when blocked or high-risk.
- Suggest improvements briefly if they are low-risk.
- Proceed with implementation by default.
- Ask for approval only if ambiguity blocks correctness.
- Keep the prompt short and focused.