ask-questions-if-underspecified

What This Skill Does

The ask-questions-if-underspecified skill acts as a critical guardrail for the OpenClaw agent, preventing the premature execution of complex tasks when requirements are ambiguous. Its primary function is to enforce a rigorous discovery phase. Instead of guessing the user's intent or proceeding with risky assumptions, the agent pauses to solicit necessary information. This skill is specifically designed to minimize "wrong work," saving time and computational resources by identifying missing context, unclear acceptance criteria, or undefined technical constraints before any code is written or files are modified.

Installation

To add this capability to your agent, execute the following command in your terminal within the OpenClaw environment:

clawhub install openclaw/skills/skills/lc0rp/ask-questions-if-underspecified

Use Cases

This skill is ideal for complex software engineering tasks or technical operations where the scope might be broad or loosely defined. It is particularly valuable for:

• Refactoring legacy codebases where implicit dependencies exist.
• Implementing new features when the business requirements lack specific edge-case handling.
• Performing infrastructure changes where rollback procedures or environment compatibility are not explicitly stated.
• Automating repetitive tasks that require specific file path patterns or naming conventions.

Example Prompts

1. "Refactor the authentication module to support OAuth2, but make sure to use this skill to clarify the existing provider constraints and database schema requirements first."
2. "I need to automate the cleanup of our log files. Use the ask-questions skill to determine which retention policies apply and whether we need to compress or archive them before deletion."
3. "Start a migration of our API documentation to the new framework, but first ask the necessary questions to ensure we aren't breaking existing cross-references."

Tips & Limitations

• Explicit Invocation: Remember that this skill does not activate automatically. You must explicitly instruct the agent to use it in your prompt.
• Keep It Lean: The skill is most effective when questions are limited to 1-5 core items. If you find the agent asking too many questions, adjust your instructions to prioritize only the "must-haves."
• Fast-Path Responses: Always define "defaults" or "recommended" settings when prompted to allow for quick iteration.
• Discovery: The skill allows for low-risk discovery, such as listing directory structures or reading config files, before final confirmation is received. Avoid allowing the agent to write files until your final confirmation step.