business-review（商业评论）

Business Review

Orchestrate an 8-agent cross-functional team to evaluate a product idea, feature, or application from multiple business stakeholder perspectives. Phase 1 runs 4 parallel strategy agents (PM, PMM, EM, Finance). Phase 2 runs 3 parallel data agents (Data Scientist, Data Engineer, Sales Engineer). Phase 3 runs a synthesizer to consolidate findings into a balanced recommendation.

Workflow

1. Parse Input

Target: $ARGUMENTS

If the target above is non-empty, use it immediately — do NOT ask the user to confirm or re-provide it. Parse it as follows:

• Feature/Product description: Extract the core idea, target users, and expected outcomes
• Context flags: Parse optional flags like --skip-data-team, --focus=sales,pm, --quick

If the target above is empty, ask the user what feature or product to evaluate and wait for their response.

Store the parsed values:

• FEATURE_DESC: The product/feature being evaluated
• FLAGS: Any processing flags (skip flags, focus filters)

IMPORTANT: When a target is provided, begin Phase 1 immediately after parsing. Do not pause for user input.

2. Phase 1 — Strategy Agents (Parallel)

Spawn 4 agents in parallel using the Task tool. Each agent is general-purpose (needs full tool access). Run all 4 with run_in_background: true for maximum parallelism.

Read REFERENCE.md first to get detailed evaluation frameworks and output templates for each agent.

Agent 1: Product Manager

subagent_type: general-purpose
run_in_background: true

Prompt:

You are a product manager evaluating the following feature/product:

{FEATURE_DESC}

Follow the "Product Manager" template in the reference below. Provide feature prioritization,
user stories, acceptance criteria, and roadmap fit analysis.

{paste Product Manager section from REFERENCE.md}

Agent 2: Product Marketing Manager

subagent_type: general-purpose
run_in_background: true

Prompt:

You are a product marketing manager evaluating the following feature/product:

{FEATURE_DESC}

Follow the "Product Marketing Manager" template in the reference below. Analyze market
positioning, competitive differentiation, messaging, and go-to-market strategy.

{paste Product Marketing Manager section from REFERENCE.md}

Agent 3: Engineering Manager

subagent_type: general-purpose
run_in_background: true

Prompt:

You are an engineering manager evaluating the following feature/product:

{FEATURE_DESC}

Follow the "Engineering Manager" template in the reference below. Assess technical feasibility,
team capacity, architecture implications, and velocity impact.

{paste Engineering Manager section from REFERENCE.md}

Agent 4: Financial Analyst

subagent_type: general-purpose
run_in_background: true

Prompt:

You are a financial analyst evaluating the following feature/product:

{FEATURE_DESC}

Follow the "Financial Analyst" template in the reference below. Analyze unit economics,
pricing strategy, revenue projections, and cost structure.

{paste Financial Analyst section from REFERENCE.md}

3. Collect Phase 1 Results

Wait for all 4 background agents to complete. Read their output files to collect results.

Compile a Phase 1 Summary containing the key findings from each agent. This summary feeds into Phase 2 agents.

4. Phase 2 — Data & Sales Agents (Parallel)

Check for --skip-data-team flag: If present, skip this entire phase and proceed to Phase 3.

Phase 2 agents run in parallel. Run all 3 with run_in_background: true.

Read REFERENCE.md for detailed templates.

Agent 5: Data Scientist

subagent_type: general-purpose
run_in_background: true

Prompt:

You are a data scientist evaluating the following feature/product:

{FEATURE_DESC}

## Phase 1 Findings
{paste compiled Phase 1 findings}

Follow the "Data Scientist" template in the reference below. Define analytics requirements,
ML opportunities, experiment design, and success metrics.

{paste Data Scientist section from REFERENCE.md}

Agent 6: Data Engineer

subagent_type: general-purpose
run_in_background: true

Prompt:

You are a data engineer evaluating the following feature/product:

{FEATURE_DESC}

## Phase 1 Findings
{paste compiled Phase 1 findings}

Follow the "Data Engineer" template in the reference below. Analyze data infrastructure needs,
pipeline design, storage strategy, and data quality requirements.

{paste Data Engineer section from REFERENCE.md}

Agent 7: Sales Engineer

subagent_type: general-purpose
run_in_background: true

Prompt:

You are a sales engineer evaluating the following feature/product:

{FEATURE_DESC}

## Phase 1 Findings
{paste compiled Phase 1 findings}

Follow the "Sales Engineer" template in the reference below. Identify customer objections,
deal blockers, demo readiness, and integration requirements.

{paste Sales Engineer section from REFERENCE.md}

5. Collect Phase 2 Results

Wait for all Phase 2 background agents to complete (or skip if --skip-data-team was set).

Compile a Phase 2 Summary with findings from data and sales agents.

6. Phase 3 — Synthesis Agent (Sequential)

Run a single synthesis agent that consumes all prior results.

subagent_type: general-purpose

Prompt:

You are a business strategist synthesizing cross-functional input on the following feature/product:

{FEATURE_DESC}

## Phase 1 Findings (Strategy)
{paste compiled Phase 1 findings}

## Phase 2 Findings (Data & Sales)
{paste compiled Phase 2 findings}

Follow the "Synthesizer" template in the reference below. Consolidate all perspectives into
a balanced recommendation with prioritized next steps.

{paste Synthesizer section from REFERENCE.md}

7. Final Report

Combine all agent outputs into a single comprehensive business review. Present to the user with this structure:

# Business Review: {FEATURE_DESC}

## Executive Summary
[3-5 bullet points: strategic value, feasibility, risks, recommendation]

## Table of Contents
1. Product Management Perspective
2. Product Marketing Perspective
3. Engineering Perspective
4. Financial Analysis
5. Data Science Perspective (if Phase 2 ran)
6. Data Engineering Perspective (if Phase 2 ran)
7. Sales Engineering Perspective (if Phase 2 ran)
8. Synthesis & Recommendation

---

## 1. Product Management Perspective

[Agent 1 output]

---

## 2. Product Marketing Perspective

[Agent 2 output]

---

## 3. Engineering Perspective

[Agent 3 output]

---

## 4. Financial Analysis

[Agent 4 output]

---

## 5. Data Science Perspective

[Agent 5 output, if Phase 2 ran]

---

## 6. Data Engineering Perspective

[Agent 6 output, if Phase 2 ran]

---

## 7. Sales Engineering Perspective

[Agent 7 output, if Phase 2 ran]

---

## 8. Synthesis & Recommendation

[Agent 8 output — Synthesizer]

---

## Decision Matrix

### GO / NO-GO Factors

| Factor | Status | Impact |
|--------|--------|--------|
| Strategic alignment | {Green/Yellow/Red} | {rationale} |
| Technical feasibility | {Green/Yellow/Red} | {rationale} |
| Financial viability | {Green/Yellow/Red} | {rationale} |
| Market opportunity | {Green/Yellow/Red} | {rationale} |
| Execution risk | {Green/Yellow/Red} | {rationale} |

### Recommended Action

**Decision**: {GO / NO-GO / CONDITIONAL GO}

**Rationale**: {2-3 sentence summary of why}

### Next Steps (Prioritized)

1. {Most critical action}
2. {Second priority}
3. {Third priority}
4. {Fourth priority}
5. {Fifth priority}

### Open Questions

1. {Critical question requiring research/decision}
2. {Second question}
3. {Third question}

Flags & Options

--skip-data-team

Skip Phase 2 entirely (data science, data engineering, sales engineering). Useful for early-stage idea validation when data infrastructure and sales engineering aren't yet relevant.

Example:

/business-review --skip-data-team Add real-time collaboration to our text editor

--focus=<roles>

Only run specific agents. Comma-separated list from: pm, pmm, em, finance, ds, de, sales, synth.

Example:

/business-review --focus=pm,em,finance Evaluate migrating to microservices

--quick

Run a fast review with reduced depth. Agents produce shorter analyses focused on critical factors only.

Example:

/business-review --quick Should we add dark mode?

Example Invocations

/business-review Add AI-powered code suggestions to our IDE
/business-review --skip-data-team Launch a community forum for users
/business-review --focus=pm,pmm,sales Evaluate entering the healthcare vertical
/business-review --quick Should we support SSO with Okta?
/business-review Build a mobile app for field technicians with offline-first architecture

Anti-Patterns

• Don't skip synthesis: The synthesizer is critical for consolidating conflicting perspectives into a coherent recommendation.
• Don't run all agents for trivial decisions: Use --quick or --focus for small features that don't warrant a full cross-functional review.
• Don't ignore red flags: If multiple agents raise concerns about the same issue (e.g., technical debt, market timing), that's a strong signal.
• Don't expect unanimous agreement: Cross-functional reviews surface trade-offs. The synthesizer's job is to balance competing priorities, not achieve consensus.
• Don't fabricate data: If agents lack information (e.g., no competitive pricing data), they should say so explicitly and recommend research rather than guessing.

Notes

• Total runtime is typically 4-10 minutes depending on complexity and whether Phase 2 runs.
• Phase 1 and Phase 2 agents run in parallel within their phases; Phase 3 runs sequentially after Phase 2.
• If an agent fails or returns thin results, note the gap in the final report rather than blocking synthesis.
• Use --focus to run only the perspectives most relevant to your decision (e.g., skip finance for internal tools).