clarify

acornsoft 的头像作者 acornsoft

Refine user intent through questions and context gathering in MacroFlow.

📖 中文导读:Refine user intent through questions and context gathering in MacroFlow. 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: clarify description: Refine user intent through questions and context gathering in MacroFlow.

Clarify Skill

Overview

The Clarify phase gathers and refines requirements, eliminating ambiguities through targeted questions and analysis. It ensures truth-seeking by probing assumptions and validating inputs. For business problem/solution development, this phase iterates on neutral Markdown from raw notes, leveraging roles (e.g., Azure Integration Specialist, Mulesoft Specialist) for reviews, analysis sessions, and ADO MCP tracking, time-boxed to 32 hours.

Purpose

  • Gather detailed requirements from raw notes and customer inputs
  • Identify and resolve ambiguities via role-based reviews (docs, code, industry updates)
  • Validate stakeholder needs and assumptions
  • Prepare neutral analysis Markdown for Plan phase

Inputs

  • Constitutional framework (principles, roles, constraints)
  • Initial requirements (raw notes, business problems)
  • Stakeholder feedback and existing assets (docs, code)

Process

  1. Gather initial inputs: Start with raw notes/customer requirements; roles review existing docs/code (e.g., Azure Solution Architect checks resources, SAP Functional Consultant audits systems).
  2. Conduct analysis sessions: Time-box to 32 hours; facilitate iterative sessions where roles build neutral Markdown (sections for findings, Mermaid diagrams); Luna agent orchestrates by prompting questions and validating via ADO work items.
  3. Resolve ambiguities: Ask targeted questions (e.g., "What integration constraints?"); leverage MCP Servers/LLMs for focused pulls; roles refine based on knowledge; agents ensure neutrality and completeness.
  4. Sync with ADO: Use MCP servers (e.g., acornsoftDevOpsServer) to query/create/link work items for findings, enabling multi-instance forensics.
  5. Establish repos/instances: Ask series of questions to add Git repos and ADO/Jira instances.
    • What Git repositories to include? (Provide URLs for GitHub/ADO repos)
    • What Azure DevOps instances/projects? (Org, project, area, iteration)
    • Any Jira instances? (URL, project)
    • Automate: Clone repos via script, configure MCP for instances.
    • Manual: Update mcp.json or run CLI commands.

Clarify Questions for Setup

  • Initial Check: "Do you need to establish any Git repositories and/or Azure DevOps instances before proceeding? (Yes/No)"
    • If Yes: Proceed to repo/instance questions; automate via Agent (Luna sub-agents) or PowerShell script, query MCP servers for validation.
  • Environment Check: "Is Visual Studio Code Insiders installed and configured? Confirm MCP servers (acornsoftDevOpsServer, ecolabDevOpsServer) are set in ~/.vscode-insiders/User/mcp.json. Is Luna Foundry setup complete (run setup-luna-foundry.ps1)?"
  • Git Repos: "List Git repository URLs to clone (e.g., https://github.com/user/repo.git)."
  • ADO Instances: "Provide ADO org/project/area/iteration (e.g., acornsoft/com.acornsoft365/com.acornsoft365(ASFT) Luna Foundry/com.acornsoft365\2026.1)."
  • Jira: "Any Jira instances? Provide URL and project key."
  • Additions: "Can add more at any time via setup script or MCP."

Outputs

  • Refined requirements document (neutral analysis Markdown)
  • Clarified scope, boundaries, and resolved ambiguities
  • ADO work item links for tracking

Grok Integration

Use Grok for probing questions and assumption validation. Temperature: 0.7 for exploratory dialogue. Luna sub-agents (e.g., @Clarify) orchestrate iterations and pulls.

Next Phase

Proceed to Specify when requirements are clear, using MacroFlow-Analysis.md as a real-world example for process improvement.

Instructions

  1. Ask targeted clarifying questions.
  2. Gather context from workspace or web.
  3. Use Grok X-insights for real-time data.
  4. Resolve assumptions.

Examples

  • Input: Vague query.
  • Output: Refined spec.

Learned Patterns

  • Parallel with Constitution.