elicitation-methodology（启发法）

Elicitation Methodology

Central hub for requirements elicitation methodology, technique selection, and workflow orchestration.

When to Use This Skill

Keywords: requirements gathering, elicitation, stakeholder needs, requirement discovery, user needs, feature requests, interview, requirements session

Invoke this skill when:

• Starting a new requirements elicitation effort
• Selecting appropriate elicitation techniques
• Orchestrating multi-source elicitation
• Configuring autonomy levels for AI assistance
• Understanding LLMREI interview patterns

Quick Decision Tree

Elicitation Techniques

1. Stakeholder Interviews (LLMREI Pattern)

AI-conducted interviews using research-backed prompting strategies.

When to use:

• Direct access to stakeholders
• Complex domains requiring exploration
• Need to capture tacit knowledge

Technique reference: See references/llmrei-patterns.md

2. Document Extraction

Mine requirements from existing documentation.

When to use:

• Existing requirements documents
• Meeting transcripts
• Regulatory documents
• Competitor analysis

Delegate to: document-extraction skill

3. Stakeholder Simulation

Multi-persona simulation for solo requirements work.

When to use:

• Working without direct stakeholder access
• Need diverse perspectives
• Validating completeness

Delegate to: stakeholder-simulation skill

4. Domain Research

MCP-powered research for domain knowledge.

When to use:

• Unfamiliar domain
• Need industry standards
• Competitive analysis
• Technology constraints

Delegate to: domain-research skill

Autonomy Levels

Guided Mode (Human-in-Loop)

autonomy: guided
behavior:
  - AI suggests questions, human approves
  - Each requirement validated individually
  - Human controls interview flow
  - Maximum transparency
use_when:
  - Sensitive or regulated domains
  - Learning the elicitation process
  - High-stakes requirements

Semi-Autonomous Mode

autonomy: semi-auto
behavior:
  - AI conducts interviews with checkpoints
  - Human validates requirement batches
  - Periodic progress reviews
  - Balance of speed and control
use_when:
  - Standard elicitation projects
  - Moderate domain complexity
  - Trusted AI capabilities

Fully Autonomous Mode

autonomy: full-auto
behavior:
  - Complete end-to-end elicitation
  - Human reviews final output only
  - Maximum efficiency
  - AI handles all decisions
use_when:
  - Well-understood domains
  - Time pressure
  - Preliminary discovery

Workflow Orchestration

Standard Discovery Workflow

1. CONTEXT GATHERING
   ├── Load any existing business context
   ├── Identify available sources (stakeholders, docs, etc.)
   └── Select autonomy level

2. MULTI-SOURCE ELICITATION
   ├── Interviews (if stakeholders available)
   ├── Document extraction (if docs available)
   ├── Domain research (MCP queries)
   └── Stakeholder simulation (if solo mode)

3. SYNTHESIS
   ├── Consolidate requirements from all sources
   ├── Remove duplicates
   ├── Classify by type (functional, NFR, constraint)
   └── Apply MoSCoW prioritization

4. VALIDATION
   ├── Gap analysis
   ├── Completeness checking
   ├── Conflict detection
   └── INVEST scoring

5. OUTPUT
   ├── Save to .requirements/{domain}/
   ├── Generate summary report
   └── Prepare for specification export

Output Format

Pre-Canonical Requirements

# .requirements/{domain}/requirements.yaml
id: REQ-SET-{number}
title: "{Domain} Requirements"
domain: "{domain-name}"
elicitation_date: "{ISO-8601-date}"
autonomy_level: "{guided|semi-auto|full-auto}"

sources:
  - type: interview|document|simulation|research
    reference: "{source-identifier}"
    timestamp: "{ISO-8601-date}"

requirements:
  - id: REQ-{number}
    text: "{requirement statement}"
    source: "{source-type}"
    source_ref: "{specific-reference}"
    priority: must|should|could|wont
    category: functional|non-functional|constraint|assumption
    confidence: high|medium|low
    validation_status: pending|validated|rejected

gaps_identified:
  - category: "{requirement-category}"
    description: "{what's missing}"
    severity: critical|major|minor

metadata:
  total_sources: {number}
  total_requirements: {number}
  gap_count: {number}
  ready_for_specification: true|false

Export Options

After elicitation, requirements can be exported to various specification formats:

/requirements-elicitation:export --to canonical  # Canonical spec format
/requirements-elicitation:export --to ears       # EARS pattern format
/requirements-elicitation:export --to gherkin    # Gherkin/BDD format

Related Skills

• interview-conducting - Detailed LLMREI interview patterns
• document-extraction - Document mining techniques
• stakeholder-simulation - Persona simulation
• gap-analysis - Completeness checking
• domain-research - MCP research coordination

References

• references/llmrei-patterns.md - LLMREI prompting strategies
• references/technique-matrix.md - Technique selection guidance
• references/autonomy-levels.md - Detailed autonomy configuration

Last Updated: 2025-12-26