ask-questions-if-underspecified

The ask-questions-if-underspecified skill is a critical tool for AI agents to ensure alignment with user intent before committing to potentially incorrect technical work. By systematically analyzing the request, the agent identifies missing details regarding project objectives, acceptance criteria, scope, technical constraints, environment requirements, or safety risks. This skill prevents the agent from starting work on the wrong path, effectively reducing the need for revisions, wasted tokens, and potential system errors caused by premature assumptions. It is specifically designed for developers, project managers, and security auditors who need to maintain high precision in AI-assisted code generation and system modification tasks. The agent acts as a guardrail, pausing operations until sufficient clarity is achieved or explicit user approval is granted.

• Performs a structured assessment of ambiguity against key criteria like objective, done definition, scope, constraints, and environment variables.

• Generates concise, scannable sets of 1-5 must-have questions to bridge knowledge gaps with minimal user friction.

• Utilizes multiple-choice formats, recommended defaults, and fast-path responses to make user interaction as efficient as possible.

• Implements a safety-first workflow: no command execution or file modification occurs until key unknowns are addressed.

• Provides a clear restatement of requirements post-clarification to ensure the agent and user are perfectly synchronized before implementation begins.

• Trigger this skill whenever a request lacks clear success criteria or carries high risk if performed incorrectly.

• Use it to enforce project constraints such as specific language versions, library dependencies, or deployment environments.

• The agent prefers low-risk discovery actions, such as inspecting configuration files (e.g., package.json, requirements.txt, or YAML files), over asking questions if the data is already available in the codebase.

• If a user requests an immediate, speculative approach, the agent will document its assumptions as a numbered list and request confirmation before proceeding, ensuring all actions are auditable and reversible.