ask-questions-if-underspecified

The ask-questions-if-underspecified skill prevents wasted effort by stopping execution when a user request contains ambiguity. Rather than guessing, the agent pauses to solicit necessary information, focusing on criteria like objectives, scope, or technical constraints. By requiring the agent to identify missing details—such as deployment targets or success definitions—before performing modifications, this skill ensures that work aligns with user intent. The agent presents options in a scannable, low-friction format, often providing recommended defaults to expedite decision-making. This disciplined approach minimizes the risk of drift or incorrect implementations that occur when instructions are vague, ensuring the final output matches project standards and safety requirements from the start.

When to Use This Skill

• •A request to refactor a codebase without specifying which components are in-scope
• •Deployment instructions that omit required environment variables or OS targets
• •Ambiguous feature requests where multiple technical implementation paths exist
• •Security-sensitive tasks where the user's risk tolerance or rollback plan is undefined

How to Invoke This Skill

Example prompts that trigger this skill in Claude Code, Cursor, or Antigravity:

• “Is the scope of this task clear enough to begin?
• “Check if you have enough information to start this refactor.
• “Are there any missing details that prevent you from starting?
• “Confirm requirements before proceeding with implementation.
• “What unknowns do you need clarified before editing these files?

Pro Tips

• 💡Always guide your AI to prioritize questions that prevent the most critical 'wrong work' first, focusing on objective and safety before minor details.
• 💡When the user is unresponsive, prompt the AI to state its assumptions and proposed path forward clearly, allowing for explicit approval or correction.
• 💡Encourage the agent to offer multiple plausible interpretations to the user, making it easier for them to identify and clarify the intended path.

What this skill does

• •Identifies vague project scope or unclear acceptance criteria in prompts
• •Constructs concise, numbered multiple-choice queries to resolve ambiguity
• •Suggests logical defaults for technical settings to accelerate user response
• •Enforces a safe-pause state until critical unknowns are resolved
• •Restates agreed-upon constraints to confirm requirements before execution

When not to use it

• ✕When a quick file-read or configuration check can resolve the missing information
• ✕When the request is already precise and contains no conflicting interpretations

Example workflow

1. Analyze the request for missing objectives, scope, or constraints.
2. Perform a low-risk scan of the repository if necessary to see if the answer is documented.
3. Formulate 1-5 targeted, numbered questions with clear default choices.
4. Present the questions to the user and request a compact reply like 'defaults' or '1b 2a'.
5. Summarize the user's choices to confirm mutual understanding.
6. Commence work once all must-have parameters are locked in.

Prerequisites

• –Access to the repository for discovery reads
• –Basic understanding of project configuration conventions

Pitfalls & limitations

• !Asking too many questions can fatigue the user; stick to the absolute must-haves.
• !Over-questioning on minor details that could be inferred from standard practice.
• !Assuming the agent knows the 'right' way to do something without verifying user preference.

FAQ

How it compares

Unlike a generic agent that might guess blindly or hallucinate requirements, this skill enforces a structured handshake, ensuring the agent and user are aligned on project constraints before a single line of code changes.

Source & trust

Recommended Rules

Recommended Workflows

Recommended MCP Servers