Loading...
Loading...
AI-led stakeholder interviews using LLMREI research-backed patterns. Conducts structured interviews to elicit requirements through context-adaptive questioning, active listening, and systematic requirement extraction.
npx skill4agent add melodic-software/claude-code-plugins interview-conductingmode: real_stakeholder
approach:
- Use AskUserQuestion tool for structured questions
- Allow natural conversation flow
- Adapt questions based on responses
- Summarize and confirm understanding periodicallymode: simulated
approach:
- Spawn persona agent via Task tool
- Conduct interview with simulated stakeholder
- Mark requirements with lower confidence
- Flag items needing real stakeholder validationStart with open-ended → Follow up with specifics → Validate understanding
Example:
Q1: "What should the system do when a user logs in?"
Q2: "You mentioned 'quick access to dashboard' - what does quick mean to you?"
Q3: "So the login should complete in under 2 seconds and show the dashboard. Is that right?"| Question | Purpose |
|---|---|
| "What is your primary goal for this system?" | High-level vision |
| "Who are the main users?" | User identification |
| "What existing systems does this replace/integrate with?" | Context mapping |
| "What would failure look like?" | Risk identification |
Pattern: [Stakeholder says X] → "When you say X, what specifically do you mean?"
Examples:
- "fast" → "What response time are you expecting? Under 1 second?"
- "secure" → "What specific security requirements apply? Authentication methods?"
- "easy to use" → "Can you describe what easy means? Any specific workflows?"Pattern: Suggest possibilities based on domain knowledge
Examples:
- "Have you considered how this works on mobile devices?"
- "What happens if the user loses connectivity mid-operation?"
- "How should the system handle peak load during [known busy period]?"requirement:
id: REQ-{number}
text: "{requirement statement}"
source: interview
stakeholder: "{role}"
timestamp: "{ISO-8601}"
type: functional|non-functional|constraint
priority: must|should|could|wont
confidence: high|medium|low
raw_quote: "{exact stakeholder words if notable}"| Mistake | Prevention |
|---|---|
| Very long questions | Keep questions concise and focused |
| Multiple unrelated questions | One question at a time |
| Leading questions | Use neutral language |
| Skipping NFRs | Explicitly ask about performance, security, usability |
| No summary | Recap periodically to verify understanding |
| Rushing | Allow silence; stakeholders often add important details |
interview_summary:
session_id: "INT-{number}"
stakeholder_role: "{role}"
duration_minutes: {number}
date: "{ISO-8601}"
autonomy_level: "{guided|semi-auto|full-auto}"
key_themes:
- "{theme-1}"
- "{theme-2}"
requirements_elicited:
- id: REQ-{number}
text: "{requirement}"
confidence: high|medium|low
type: functional|non-functional|constraint
priority: must|should|could
follow_up_needed:
- "{question or topic needing clarification}"
stakeholder_quotes:
- "{notable direct quote}"
observations:
- "{interviewer observation about needs or concerns}"
next_steps:
- "{recommended action}"references/llmrei-patterns.mdstakeholder-simulationdomain-research.requirements/{domain}/interviews/INT-{number}.yamlelicitation-methodologystakeholder-simulationgap-analysis