playbooks

home / skills / multiversx / mx-ai-skills / clarification_expert

clarification_expert skill

/antigravity/skills/clarification_expert

This skill identifies ambiguities in user requests and prompts targeted clarifying questions to unblock design, development, or auditing tasks.

npx playbooks add skill multiversx/mx-ai-skills --skill clarification_expert

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
1.4 KB
---
name: clarification_expert
description: Expert at identifying underspecified requirements and asking high-value clarifying questions.
---

# Clarification Expert

This skill helps you identify ambiguity in user requests and ask targeted questions to unblock development or auditing tasks.

## When to Use
- The user's request is vague (e.g., "Make it secure").
- Missing technical constraints (e.g., "Add a token" but no standard specified).
- Conflicting requirements.

## Guidelines for Questions

### 1. Be Specific
Don't ask "What do you mean?". Ask "Do you want X or Y?".
*   *Bad*: "How should the token work?"
*   *Good*: "Should the token be a standard Fungible ESDT or a Semi-Fungible SFT with metadata?"

### 2. Batch Questions
Don't ask one question at a time. Group related questions into a numbered list.

### 3. Propose Defaults
Always offer a sensible default if the user doesn't know.
*   *Example*: "If you don't have a preference, I recommend using `SingleValueMapper` for the config storage to save gas. Shall I proceed with that?"

## Analysis Categories

1.  **Scope**: Is it a full audit or just a specific module?
2.  **Environment**: Mainnet, Devnet, or Sovereign Chain?
3.  **Risk Profile**: Is this a generic dApp or a high-value DeFi protocol?
4.  **Tech Stack**: Are we using standard `multiversx-sc` modules or custom unchecked arithmetic? (Rule: Always suggest standard modules).

Overview

This skill identifies underspecified requirements and generates targeted clarifying questions to unblock development, audits, and product decisions. It focuses on converting vague statements into actionable choices and sensible defaults. The goal is faster, safer progress by reducing assumptions and rework.

How this skill works

The skill inspects a request for missing constraints, ambiguities, and conflicting goals across scope, environment, risk profile, and tech stack. It outputs a numbered list of high-value questions grouped by topic and proposes a recommended default when the requester lacks a preference. It also highlights risky assumptions and suggests standard libraries or modules where appropriate.

When to use it

  • Request is vague or open-ended (e.g., “Make it secure”).
  • Requirements are missing technical constraints (tokens, storage, performance).
  • Requirements conflict or imply trade-offs (security vs gas vs UX).
  • Preparing for audits, deployments, or cross-team handoffs.
  • Designing integrations that depend on environment or chain specifics.

Best practices

  • Ask specific, actionable questions — avoid “What do you mean?”; offer concrete options instead.
  • Batch related questions into a numbered, grouped list to avoid endless back-and-forth.
  • Always propose a sensible default and note why it’s recommended.
  • Call out the analysis categories: Scope, Environment, Risk Profile, Tech Stack.
  • Recommend standard, audited modules where possible; flag custom unchecked approaches as higher risk.

Example use cases

  • A product manager says “Add a token” — produce questions about token standard, supply, minting, metadata, and propose a default standard.
  • An engineer requests “Make this secure” — clarify threat model, attacker capabilities, required certifications, and propose practical mitigation priorities.
  • Before an audit, create a scope checklist: full system vs module, target environment, and high-risk components to prioritize.
  • Integrating third-party services — clarify expected latency, failure modes, and whether retries or fallbacks are acceptable.

FAQ

What makes a question "high-value"?

High-value questions remove major unknowns that change design or risk profile, such as environment, attacker model, or required standards.

What if the requester truly has no preference?

Provide a recommended default with rationale and state how to change it later; defaults speed progress while keeping options reversible.