Customer Support — Customer Support System Building Pipeline
An agent team collaborates to generate FAQ, response manual, escalation policy, and analytics all at once.
Execution Mode
Agent Team — 5 members communicate directly via SendMessage and perform cross-validation.
Agent Composition
| Agent |
File |
Role |
Type |
| faq-builder |
.claude/agents/faq-builder.md |
FAQ construction, search optimization |
general-purpose |
| response-specialist |
.claude/agents/response-specialist.md |
Response scripts, tone & manner |
general-purpose |
| escalation-manager |
.claude/agents/escalation-manager.md |
Escalation tiers, SLA |
general-purpose |
| cs-analyst |
.claude/agents/cs-analyst.md |
Metrics, VOC, improvement proposals |
general-purpose |
| cs-reviewer |
.claude/agents/cs-reviewer.md |
Cross-validation, customer journey simulation |
general-purpose |
Workflow
Phase 1: Preparation (Performed directly by orchestrator)
- Extract from user input:
- Service Information: Product/service name, type, customer base size
- Support Channels: Which channels are in use — phone/chat/email/social media
- Current State: Whether an existing CS system exists, team size
- Existing Materials (optional): Existing FAQs, manuals, data
- Create
_workspace/ directory at the project root
- Organize the input and save to
_workspace/00_input.md
- If existing files are provided, copy them to
_workspace/ and skip the corresponding Phase
Phase 2: Team Assembly and Execution
| Order |
Task |
Assigned To |
Dependencies |
Deliverable |
| 1a |
FAQ Construction |
faq-builder |
None |
_workspace/01_faq.md |
| 1b |
Response Manual |
response-specialist |
None |
_workspace/02_response_manual.md |
| 2 |
Escalation Policy |
escalation-manager |
Tasks 1a, 1b |
_workspace/03_escalation_policy.md |
| 3 |
CS Analytics Framework |
cs-analyst |
Tasks 1a, 1b, 2 |
_workspace/04_cs_analytics.md |
| 4 |
CS Review |
cs-reviewer |
Tasks 1a, 1b, 2, 3 |
_workspace/05_review_report.md |
Tasks 1a (FAQ) and 1b (Manual) are executed in parallel. Both can start simultaneously as they have no initial dependencies.
Inter-team Communication Flow:
- faq-builder completes → delivers out-of-FAQ-scope scenarios to response-specialist, delivers self-service boundary to escalation-manager
- response-specialist completes → delivers escalation trigger conditions to escalation-manager
- escalation-manager completes → delivers SLA metrics to cs-analyst
- cs-analyst completes → delivers all documents to cs-reviewer
- cs-reviewer cross-validates all deliverables. If 🔴 Must Fix items are found, sends revision requests to the relevant agent → rework → re-validate (up to 2 times)
Phase 3: Integration and Final Deliverables
- Verify all files in
_workspace/
- Confirm that all 🔴 Must Fix items from the review report have been addressed
- Report the final summary to the user
Execution Modes by Request Scale
| User Request Pattern |
Execution Mode |
Agents Deployed |
| "Build a customer support system", "Full CS setup" |
Full Pipeline |
All 5 |
| "Just create FAQs" |
FAQ Mode |
faq-builder + reviewer |
| "Write a response manual" |
Manual Mode |
response-specialist + reviewer |
| "Create an escalation policy" |
Escalation Mode |
escalation-manager + reviewer |
| "Design CS metrics" |
Analytics Mode |
cs-analyst + reviewer |
| "Review this FAQ" |
Review Mode |
reviewer only |
Data Transfer Protocol
| Strategy |
Method |
Purpose |
| File-based |
_workspace/ directory |
Storing and sharing main deliverables |
| Message-based |
SendMessage |
Real-time key information transfer, revision requests |
Error Handling
| Error Type |
Strategy |
| Insufficient service info |
Build a generic CS system based on industry type, tag as "requires service customization" |
| No existing CS data |
Set targets based on industry benchmarks, propose initial measurement methods |
| Web search failure |
Work from general CS best practices, note "data limited" |
| Agent failure |
Retry once → if still fails, proceed without that deliverable, note the omission in the review report |
| 🔴 found in review |
Send revision request to the relevant agent → rework → re-validate (up to 2 times) |
Test Scenarios
Normal Flow
Prompt: "Build a customer support system for a B2C SaaS project management tool. We have 10,000 monthly active users and operate chat and email channels. The CS team has 5 members."
Expected Results:
- FAQ: Sign-up/billing/features/integration categories, 30+ Q&As
- Manual: 5+ scenario scripts, channel-specific guides for chat + email
- Escalation: L1-L3 tiers, SLA matrix, crisis response protocol
- Analytics: CSAT/NPS/FCR/AHT metrics + weekly/monthly report templates
- Review: Includes 3 customer journey simulations
Existing File Utilization Flow
Prompt: "I have this FAQ; create a response manual and escalation policy" + FAQ file attached
Expected Results:
- Copy existing FAQ to
_workspace/01_faq.md
- Skip faq-builder and deploy response-specialist + escalation-manager + cs-analyst + reviewer
Error Flow
Prompt: "Create a customer support FAQ"
Expected Results:
- Switch to FAQ Mode (faq-builder + reviewer)
- If service info is insufficient, ask follow-up questions about industry/product type
- If only minimal info is provided, generate a generic SaaS FAQ template
Agent Extension Skills
Extension skills that enhance agent domain expertise:
| Skill |
File |
Target Agents |
Role |
| csat-analyzer |
.claude/skills/csat-analyzer/skill.md |
cs-analyst, cs-reviewer |
CSAT/NPS/CES design, operational metrics, VOC analysis, CS dashboard design |
| escalation-flowchart |
.claude/skills/escalation-flowchart/skill.md |
escalation-manager, response-specialist |
L1/L2/L3 structure, severity classification, trigger conditions, SLA matrix, crisis response |
1---2name: customer-support3description: A full pipeline that systematizes a customer support system into FAQ, response manual, escalation policy, and CS analytics framework. Use this skill for requests like 'build a customer support system', 'write FAQs', 'response manual', 'CS manual', 'escalation policy', 'customer response scripts', 'complaint handling', 'customer complaint processing', 'CS operations system', 'customer service process', and other customer support system building needs. Also supports supplementing and improving existing FAQs or manuals. However, CRM/helpdesk software development, call center infrastructure setup, real-time chatbot development, and agent recruitment are outside the scope of this skill.4---56# Customer Support — Customer Support System Building Pipeline78An agent team collaborates to generate FAQ, response manual, escalation policy, and analytics all at once.910## Execution Mode1112**Agent Team** — 5 members communicate directly via SendMessage and perform cross-validation.1314## Agent Composition1516| Agent | File | Role | Type |17|-------|------|------|------|18| faq-builder | `.claude/agents/faq-builder.md` | FAQ construction, search optimization | general-purpose |19| response-specialist | `.claude/agents/response-specialist.md` | Response scripts, tone & manner | general-purpose |20| escalation-manager | `.claude/agents/escalation-manager.md` | Escalation tiers, SLA | general-purpose |21| cs-analyst | `.claude/agents/cs-analyst.md` | Metrics, VOC, improvement proposals | general-purpose |22| cs-reviewer | `.claude/agents/cs-reviewer.md` | Cross-validation, customer journey simulation | general-purpose |2324## Workflow2526### Phase 1: Preparation (Performed directly by orchestrator)27281. Extract from user input:29 - **Service Information**: Product/service name, type, customer base size30 - **Support Channels**: Which channels are in use — phone/chat/email/social media31 - **Current State**: Whether an existing CS system exists, team size32 - **Existing Materials** (optional): Existing FAQs, manuals, data332. Create `_workspace/` directory at the project root343. Organize the input and save to `_workspace/00_input.md`354. If existing files are provided, copy them to `_workspace/` and skip the corresponding Phase3637### Phase 2: Team Assembly and Execution3839| Order | Task | Assigned To | Dependencies | Deliverable |40|-------|------|-------------|-------------|-------------|41| 1a | FAQ Construction | faq-builder | None | `_workspace/01_faq.md` |42| 1b | Response Manual | response-specialist | None | `_workspace/02_response_manual.md` |43| 2 | Escalation Policy | escalation-manager | Tasks 1a, 1b | `_workspace/03_escalation_policy.md` |44| 3 | CS Analytics Framework | cs-analyst | Tasks 1a, 1b, 2 | `_workspace/04_cs_analytics.md` |45| 4 | CS Review | cs-reviewer | Tasks 1a, 1b, 2, 3 | `_workspace/05_review_report.md` |4647Tasks 1a (FAQ) and 1b (Manual) are **executed in parallel**. Both can start simultaneously as they have no initial dependencies.4849**Inter-team Communication Flow:**50- faq-builder completes → delivers out-of-FAQ-scope scenarios to response-specialist, delivers self-service boundary to escalation-manager51- response-specialist completes → delivers escalation trigger conditions to escalation-manager52- escalation-manager completes → delivers SLA metrics to cs-analyst53- cs-analyst completes → delivers all documents to cs-reviewer54- cs-reviewer cross-validates all deliverables. If 🔴 Must Fix items are found, sends revision requests to the relevant agent → rework → re-validate (up to 2 times)5556### Phase 3: Integration and Final Deliverables57581. Verify all files in `_workspace/`592. Confirm that all 🔴 Must Fix items from the review report have been addressed603. Report the final summary to the user6162## Execution Modes by Request Scale6364| User Request Pattern | Execution Mode | Agents Deployed |65|---------------------|----------------|-----------------|66| "Build a customer support system", "Full CS setup" | **Full Pipeline** | All 5 |67| "Just create FAQs" | **FAQ Mode** | faq-builder + reviewer |68| "Write a response manual" | **Manual Mode** | response-specialist + reviewer |69| "Create an escalation policy" | **Escalation Mode** | escalation-manager + reviewer |70| "Design CS metrics" | **Analytics Mode** | cs-analyst + reviewer |71| "Review this FAQ" | **Review Mode** | reviewer only |7273## Data Transfer Protocol7475| Strategy | Method | Purpose |76|----------|--------|---------|77| File-based | `_workspace/` directory | Storing and sharing main deliverables |78| Message-based | SendMessage | Real-time key information transfer, revision requests |7980## Error Handling8182| Error Type | Strategy |83|-----------|----------|84| Insufficient service info | Build a generic CS system based on industry type, tag as "requires service customization" |85| No existing CS data | Set targets based on industry benchmarks, propose initial measurement methods |86| Web search failure | Work from general CS best practices, note "data limited" |87| Agent failure | Retry once → if still fails, proceed without that deliverable, note the omission in the review report |88| 🔴 found in review | Send revision request to the relevant agent → rework → re-validate (up to 2 times) |8990## Test Scenarios9192### Normal Flow93**Prompt**: "Build a customer support system for a B2C SaaS project management tool. We have 10,000 monthly active users and operate chat and email channels. The CS team has 5 members."94**Expected Results**:95- FAQ: Sign-up/billing/features/integration categories, 30+ Q&As96- Manual: 5+ scenario scripts, channel-specific guides for chat + email97- Escalation: L1-L3 tiers, SLA matrix, crisis response protocol98- Analytics: CSAT/NPS/FCR/AHT metrics + weekly/monthly report templates99- Review: Includes 3 customer journey simulations100101### Existing File Utilization Flow102**Prompt**: "I have this FAQ; create a response manual and escalation policy" + FAQ file attached103**Expected Results**:104- Copy existing FAQ to `_workspace/01_faq.md`105- Skip faq-builder and deploy response-specialist + escalation-manager + cs-analyst + reviewer106107### Error Flow108**Prompt**: "Create a customer support FAQ"109**Expected Results**:110- Switch to FAQ Mode (faq-builder + reviewer)111- If service info is insufficient, ask follow-up questions about industry/product type112- If only minimal info is provided, generate a generic SaaS FAQ template113114## Agent Extension Skills115116Extension skills that enhance agent domain expertise:117118| Skill | File | Target Agents | Role |119|-------|------|---------------|------|120| csat-analyzer | `.claude/skills/csat-analyzer/skill.md` | cs-analyst, cs-reviewer | CSAT/NPS/CES design, operational metrics, VOC analysis, CS dashboard design |121| escalation-flowchart | `.claude/skills/escalation-flowchart/skill.md` | escalation-manager, response-specialist | L1/L2/L3 structure, severity classification, trigger conditions, SLA matrix, crisis response |
revfactory