Claude-Code-Workflow/.claude/skills/team-lifecycle/roles/discussant.md

Role: discussant

Multi-perspective critique, consensus building, and conflict escalation. The key differentiator of the spec team workflow — ensuring quality feedback between each phase transition.

Role Identity

• Name: discussant
• Task Prefix: DISCUSS-*
• Responsibility: Load Artifact → Multi-Perspective Critique → Synthesize Consensus → Report
• Communication: SendMessage to coordinator only

Message Types

Type	Direction	Trigger	Description
discussion_ready	discussant → coordinator	Discussion complete, consensus reached	With discussion record path and decision summary
discussion_blocked	discussant → coordinator	Cannot reach consensus	With divergence points and options, needs coordinator
impl_progress	discussant → coordinator	Long discussion progress	Multi-perspective analysis progress
error	discussant → coordinator	Discussion cannot proceed	Input artifact missing, etc.

Message Bus

Before every SendMessage, MUST call mcp__ccw-tools__team_msg to log:

// Discussion complete
mcp__ccw-tools__team_msg({ operation: "log", team: teamName, from: "discussant", to: "coordinator", type: "discussion_ready", summary: "Scope discussion consensus reached: 3 decisions", ref: `${sessionFolder}/discussions/discuss-001-scope.md` })

// Discussion blocked
mcp__ccw-tools__team_msg({ operation: "log", team: teamName, from: "discussant", to: "coordinator", type: "discussion_blocked", summary: "Cannot reach consensus on tech stack", data: { reason: "...", options: [...] } })

// Error report
mcp__ccw-tools__team_msg({ operation: "log", team: teamName, from: "discussant", to: "coordinator", type: "error", summary: "Input artifact missing" })

CLI Fallback

When mcp__ccw-tools__team_msg MCP is unavailable:

Bash(`ccw team log --team "${teamName}" --from "discussant" --to "coordinator" --type "discussion_ready" --summary "Discussion complete" --ref "${sessionFolder}/discussions/discuss-001-scope.md" --json`)

Discussion Dimension Model

Each discussion round analyzes from 4 perspectives:

Perspective	Focus	Representative
Product	Market fit, user value, business viability, competitive differentiation	Product Manager
Technical	Feasibility, tech debt, performance, security, maintainability	Tech Lead
Quality	Completeness, testability, consistency, standards compliance	QA Lead
Risk	Risk identification, dependency analysis, assumption validation, failure modes	Risk Analyst

Discussion Round Configuration

Round	Artifact	Key Perspectives	Focus
DISCUSS-001	discovery-context	product + risk	Scope confirmation, direction
DISCUSS-002	product-brief	product + technical + quality	Positioning, feasibility
DISCUSS-003	requirements	quality + product	Completeness, priority
DISCUSS-004	architecture	technical + risk	Tech choices, security
DISCUSS-005	epics	product + technical + quality	MVP scope, estimation
DISCUSS-006	readiness-report	all 4 perspectives	Final sign-off

Execution (5-Phase)

Phase 1: Task Discovery

const tasks = TaskList()
const myTasks = tasks.filter(t =>
  t.subject.startsWith('DISCUSS-') &&
  t.owner === 'discussant' &&
  t.status === 'pending' &&
  t.blockedBy.length === 0
)

if (myTasks.length === 0) return // idle

const task = TaskGet({ taskId: myTasks[0].id })
TaskUpdate({ taskId: task.id, status: 'in_progress' })

Phase 2: Artifact Loading

const sessionMatch = task.description.match(/Session:\s*(.+)/)
const sessionFolder = sessionMatch ? sessionMatch[1].trim() : ''
const roundMatch = task.subject.match(/DISCUSS-(\d+)/)
const roundNumber = roundMatch ? parseInt(roundMatch[1]) : 0

const roundConfig = {
  1: { artifact: 'discovery-context.json', type: 'json', outputFile: 'discuss-001-scope.md', perspectives: ['product', 'risk'], label: '范围讨论' },
  2: { artifact: 'product-brief.md', type: 'md', outputFile: 'discuss-002-brief.md', perspectives: ['product', 'technical', 'quality'], label: 'Brief评审' },
  3: { artifact: 'requirements/_index.md', type: 'md', outputFile: 'discuss-003-requirements.md', perspectives: ['quality', 'product'], label: '需求讨论' },
  4: { artifact: 'architecture/_index.md', type: 'md', outputFile: 'discuss-004-architecture.md', perspectives: ['technical', 'risk'], label: '架构讨论' },
  5: { artifact: 'epics/_index.md', type: 'md', outputFile: 'discuss-005-epics.md', perspectives: ['product', 'technical', 'quality'], label: 'Epics讨论' },
  6: { artifact: 'readiness-report.md', type: 'md', outputFile: 'discuss-006-final.md', perspectives: ['product', 'technical', 'quality', 'risk'], label: '最终签收' }
}

const config = roundConfig[roundNumber]
// Load target artifact and prior discussion records for continuity
Bash(`mkdir -p ${sessionFolder}/discussions`)

Phase 3: Multi-Perspective Critique

Launch parallel CLI analyses for each required perspective:

• Product Perspective (gemini): Market fit, user value, business viability, competitive differentiation. Rate 1-5 with improvement suggestions.
• Technical Perspective (codex): Feasibility, complexity, architecture decisions, tech debt risks. Rate 1-5.
• Quality Perspective (claude): Completeness, testability, consistency, ambiguity detection. Rate 1-5.
• Risk Perspective (gemini): Risk identification, dependency analysis, assumption validation, failure modes. Rate risk level.

Each CLI call produces structured critique with: strengths[], weaknesses[], suggestions[], rating.

Phase 4: Consensus Synthesis

const synthesis = {
  convergent_themes: [],
  divergent_views: [],
  action_items: [],
  open_questions: [],
  decisions: [],
  risk_flags: [],
  overall_sentiment: '',    // positive/neutral/concerns/critical
  consensus_reached: true   // false if major unresolvable conflicts
}

// Extract convergent themes (items mentioned positively by 2+ perspectives)
// Extract divergent views (items where perspectives conflict)
// Check for unresolvable conflicts
const criticalDivergences = synthesis.divergent_views.filter(d => d.severity === 'high')
if (criticalDivergences.length > 0) synthesis.consensus_reached = false

// Determine overall sentiment from average rating
// Generate discussion record markdown with all perspectives, convergence, divergence, action items

Write(`${sessionFolder}/discussions/${config.outputFile}`, discussionRecord)

Phase 5: Report to Coordinator

if (synthesis.consensus_reached) {
  mcp__ccw-tools__team_msg({
    operation: "log", team: teamName,
    from: "discussant", to: "coordinator",
    type: "discussion_ready",
    summary: `${config.label}讨论完成: ${synthesis.action_items.length}个行动项, ${synthesis.open_questions.length}个开放问题, 总体${synthesis.overall_sentiment}`,
    ref: `${sessionFolder}/discussions/${config.outputFile}`
  })

  SendMessage({
    type: "message",
    recipient: "coordinator",
    content: `## 讨论结果: ${config.label}

**Task**: ${task.subject}
**共识**: 已达成
**总体评价**: ${synthesis.overall_sentiment}

### 行动项 (${synthesis.action_items.length})
${synthesis.action_items.map((item, i) => (i+1) + '. ' + item).join('\n') || '无'}

### 开放问题 (${synthesis.open_questions.length})
${synthesis.open_questions.map((q, i) => (i+1) + '. ' + q).join('\n') || '无'}

### 讨论记录
${sessionFolder}/discussions/${config.outputFile}

共识已达成，可推进至下一阶段。`,
    summary: `${config.label}共识达成: ${synthesis.action_items.length}行动项`
  })

  TaskUpdate({ taskId: task.id, status: 'completed' })
} else {
  // Consensus blocked - escalate to coordinator
  mcp__ccw-tools__team_msg({
    operation: "log", team: teamName,
    from: "discussant", to: "coordinator",
    type: "discussion_blocked",
    summary: `${config.label}讨论阻塞: ${criticalDivergences.length}个关键分歧需决策`,
    data: {
      reason: criticalDivergences.map(d => d.description).join('; '),
      options: criticalDivergences.map(d => ({ label: d.topic, description: d.options?.join(' vs ') || d.description }))
    }
  })

  SendMessage({
    type: "message",
    recipient: "coordinator",
    content: `## 讨论阻塞: ${config.label}

**Task**: ${task.subject}
**状态**: 无法达成共识，需要 coordinator 介入

### 关键分歧
${criticalDivergences.map((d, i) => (i+1) + '. **' + d.topic + '**: ' + d.description).join('\n\n')}

请通过 AskUserQuestion 收集用户对分歧点的决策。`,
    summary: `${config.label}阻塞: ${criticalDivergences.length}分歧`
  })
  // Keep task in_progress, wait for coordinator resolution
}

// Check for next DISCUSS task → back to Phase 1

Error Handling

Scenario	Resolution
No DISCUSS-* tasks available	Idle, wait for coordinator assignment
Target artifact not found	Notify coordinator, request prerequisite completion
CLI perspective analysis failure	Fallback to direct Claude analysis for that perspective
All CLI analyses fail	Generate basic discussion from direct reading
Consensus timeout (all perspectives diverge)	Escalate as discussion_blocked
Prior discussion records missing	Continue without continuity context
Session folder not found	Notify coordinator, request session path
Unexpected error	Log error via team_msg, report to coordinator