ask-questions-if-underspecified

Asks targeted clarifying questions before implementing ambiguous requests.

Updated May 17, 2026
One-click install
npx skills add https://github.com/irrit-us/agent_misc --skill ask-questions-if-underspecified-irrit-us
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ask-questions-if-underspecified
Source: https://github.com/irrit-us/agent_misc/tree/main/skills/ask-questions-if-underspecified
Command: npx skills add https://github.com/irrit-us/agent_misc --skill ask-questions-if-underspecified-irrit-us

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Ambiguous or underspecified requests often lead to wasted work, wrong implementations, and costly rework. This Skill ensures the AI pauses to clarify objectives, scope, constraints, and acceptance criteria before writing any code or making changes. ## Core Features & Use Cases - Underspecification Detection: Evaluates requests against a checklist covering objective, definition of done, scope, constraints, environment, and safety to decide whether clarification is needed. - Structured Questioning: Asks 1-5 must-have questions with multiple-choice options, recommended defaults, and compact reply formats (e.g., 1a 2b) to minimize user effort. - Safe Pause Behavior: Blocks commands, edits, and detailed plans until answers arrive, while permitting clearly labeled low-risk discovery reads. - Use Case: A user asks to "improve the API." Instead of guessing, the AI asks which endpoints are in scope, what performance target defines done, and whether breaking changes are acceptable, then confirms the interpretation before starting. ## Quick Start Ask the AI to review your request for ambiguity and ask clarifying questions before implementing anything.

Frequently Asked Questions about ask-questions-if-underspecified

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get an AI to ask clarifying questions before coding?▼

Use a clarification workflow that detects underspecified requests and asks 1-5 must-have questions before any implementation. The AI pauses edits and commands until answers arrive, then restates the requirements to confirm understanding.

When should an AI ask questions instead of proceeding?▼

Ask questions when multiple plausible interpretations exist or key details like objective, scope, constraints, environment, or safety are unclear. Skip questions when the request is already clear or a quick low-risk read of configs and docs can answer them.

How many clarifying questions should be asked at once?▼

Ask 1-5 questions in the first pass, prioritizing ones that eliminate whole branches of work. Use numbered questions with lettered multiple-choice options, marked defaults, and a compact reply format like "1a 2b" to reduce friction.

What are the limitations of asking clarifying questions?▼

Over-asking slows down simple tasks and frustrates users when answers are discoverable from the repository. Avoid open-ended questions when a tight multiple-choice or yes/no would resolve ambiguity faster, and never ask about things a quick discovery read can answer.

What happens if the user says to proceed without answering?▼

The AI states its assumptions as a short numbered list and asks for confirmation. It proceeds only after the user confirms or corrects those assumptions, ensuring the work still rests on an explicit shared understanding.