shield_lock
金大哥
← 返回 Skill 列表
extension
分类: 营销与增长无需 API Key

product-ideation

通过一个六阶段的研究流程评估产品想法，该流程基于市场分析、竞争情报和细分市场定位生成买入/持有/卖出建议。在评估产品可行性、为新想法研究市场或决定是否追求某个产品概念时使用。

person作者: jakexiaohubgithub
open_in_new仓库download下载资源包

Product Ideation

Evaluates a product idea through a structured six-agent research pipeline and delivers a BUY/HOLD/SELL recommendation. The orchestrator interviews the user to clarify the idea, then spawns dedicated sub-agents for validation, market research, competitive analysis, pattern documentation, segment analysis, and final strategy. All analysis is grounded in web research and structured frameworks.

When to Use This Skill

Load this skill when the user request matches ANY of these patterns:

| Trigger Pattern | Example User Request | |-----------------|----------------------| | Idea evaluation | "Is this product idea worth pursuing?", "Evaluate my idea: X" | | Market analysis | "Research the market for X", "Should I build X?" | | Investment-style assessment | "Give me a BUY/HOLD/SELL on this idea", "Validate my product concept" | | Pre-build research | "Before I start building X, help me understand the landscape" |

DO NOT use for:

  • Evaluating existing live products or companies (use a financial analysis skill instead)
  • Technical architecture decisions for an already-committed product (the idea stage has passed)
  • Competitive analysis in isolation without a product idea as the anchor

Dependencies

| Category | Files | Requirement | When to Load | |----------|-------|-------------|--------------| | Analysis frameworks | references/analysis-frameworks.md | REQUIRED | Include in sub-agent prompts at Stage 1+ | | Final report template | templates/final-report.md | REQUIRED | Include in strategist sub-agent prompt | | Diagnostics schema | templates/diagnostic-output.yaml | REQUIRED | Write at pipeline completion | | Prompting | subagent-prompting skill | REQUIRED | Load before spawning any sub-agent | | Idea validator | .claude/agents/product-ideation-idea-validator.md | REQUIRED | Spawned at Stage 1 (parallel) | | Market researcher | .claude/agents/product-ideation-market-researcher.md | REQUIRED | Spawned at Stage 1 (parallel) | | Competitive analyzer | .claude/agents/product-ideation-competitive-analyzer.md | REQUIRED | Spawned at Stage 2 (sequential) | | Pattern documenter | .claude/agents/product-ideation-pattern-documenter.md | REQUIRED | Spawned at Stage 3 (parallel) | | Segment analyzer | .claude/agents/product-ideation-segment-analyzer.md | REQUIRED | Spawned at Stage 3 (parallel) | | Strategist | .claude/agents/product-ideation-strategist.md | REQUIRED | Spawned at Stage 4 (sequential) |

Usage

/product-ideation <idea-description>
/product-ideation --doc <path-to-idea-file>

Arguments:

  • <idea-description> — Free-text description of the product idea
  • --doc <path> — Path to a file containing the idea description

Examples:

  • /product-ideation a mobile app that lets freelancers track client payments with automated invoice reminders
  • /product-ideation --doc plans/ideas/saas-invoicing-tool.md

Pre-Flight Gate (BLOCKING)

STOP. Before ANY work, you MUST acknowledge what this skill requires.

This skill uses a six-agent pipeline with parallel and sequential stages. You are the orchestrator. You coordinate, read outputs, and spawn agents — you do NOT perform their analysis yourself.

What You MUST Do

  1. Parse input and conduct a focused interview (Stage 0) — at most 2 AskUserQuestion rounds
  2. Read references/analysis-frameworks.md and templates/final-report.md before spawning agents
  3. Spawn Stage 1 agents (idea-validator AND market-researcher) in the same message as parallel Tasks
  4. Wait for both Stage 1 agents to complete before spawning Stage 2
  5. Spawn Stage 3 agents (pattern-documenter AND segment-analyzer) in the same message as parallel Tasks
  6. Wait for all prior outputs before spawning the strategist
  7. Read the final strategy report from $PROJECT_DIR/logs/ and present its contents to the user
  8. Write diagnostic YAML at completion

What You MUST NOT Do

  • Do NOT perform market research, validation, or competitive analysis yourself — spawn the agents
  • Do NOT skip the interview — the idea must be clear before spawning agents
  • Do NOT spawn agents with run_in_background: true
  • Do NOT proceed past Stage 1 without confirming both parallel agents completed
  • Do NOT return to user before the strategist has written its output to $PROJECT_DIR/logs/

Anti-thought traps:

  • "I already know this market, I can skip the research agents" — STOP. Spawn the agents.
  • "The idea is simple, I don't need to interview" — STOP. Conduct at least one AskUserQuestion round.
  • "I can summarize the recommendation without reading the strategist output" — STOP. Read the log file.

Pipeline

// product-ideation pipeline
Stage0_Interview(args)
|> [Stage1_IdeaValidator(idea),         // product-ideation-idea-validator — feasibility + timing
    Stage1_MarketResearcher(idea)]       // product-ideation-market-researcher — market size + trends
|> Stage2_CompetitiveAnalyzer(market_research_output)  // product-ideation-competitive-analyzer
|> [Stage3_PatternDocumenter(competitive_output),       // product-ideation-pattern-documenter
    Stage3_SegmentAnalyzer(competitive_output)]         // product-ideation-segment-analyzer
|> Stage4_Strategist(all_outputs)       // product-ideation-strategist — BUY/HOLD/SELL
|> Stage5_Diagnostics(pipeline_results)

Stage Definitions

Stage 0: Interview (Orchestrator)

Stage 0: Interview
├── Parse arguments:
│   ├── If --doc <path>: Read the file, extract idea description
│   └── If free-text: Use as idea description directly
├── Load references/analysis-frameworks.md
├── Load templates/final-report.md
├── AskUserQuestion (Round 1 — REQUIRED):
│   ├── Q1: What problem does this product solve, and for whom?
│   ├── Q2: What is your rough idea of the business model (subscription, one-time, marketplace, etc.)?
│   ├── Q3: Are there any specific markets or geographies to focus on?
│   └── Q4: Do you have any known competitors in mind, or is this a blank slate?
├── If answers are ambiguous or incomplete:
│   └── AskUserQuestion (Round 2 — CONDITIONAL, max 1 additional round):
│       └── Ask 1-2 targeted follow-ups based on gaps in Round 1 answers
├── Synthesize: Combine original idea + interview answers into a complete idea brief
└── Token budget check (warn if >20% consumed before agents start)

Interview behavior: Maximum 2 AskUserQuestion rounds. Present all Round 1 questions together. Do NOT ask questions one at a time.

Stage 1: Parallel Batch 1 — Idea Validator + Market Researcher

Spawn both agents in the same message. Do NOT wait for one before spawning the other.

Idea Validator prompt structure:

GOAL: Assess whether the product idea has sufficient merit to warrant deep market research.
Evaluate technical feasibility, market timing, problem-solution fit, and uniqueness.

CONSTRAINTS:
- Use WebSearch to check for prior art and existing solutions
- Produce a clear PASS/CONDITIONAL/FAIL verdict with reasoning
- Write output to $PROJECT_DIR/logs/product-ideation-idea-validation-{YYYYMMDD-HHMMSS}.md
- Do not exceed 500 words in the main report

CONTEXT:
Idea brief: {synthesized idea brief from Stage 0}
Frameworks reference: {path to references/analysis-frameworks.md}

OUTPUT:
Write report to $PROJECT_DIR/logs/product-ideation-idea-validation-{YYYYMMDD-HHMMSS}.md
Return summary to orchestrator: verdict, top 3 strengths, top 3 concerns, report path

Spawn: Task(subagent_type="product-ideation-idea-validator", prompt=...)

Market Researcher prompt structure:

GOAL: Research the market landscape for this product idea. Produce an evidence-based
assessment of market size, growth trajectory, key players, regulatory context, and
technology trends. Ground all claims in web research.

CONSTRAINTS:
- Use WebSearch and WebFetch extensively — minimum 5 distinct sources
- Quantify market size where possible (TAM, growth rate, CAGR)
- Note recency of data — prefer sources from 2023 or later
- Write output to $PROJECT_DIR/logs/product-ideation-market-research-{YYYYMMDD-HHMMSS}.md

CONTEXT:
Idea brief: {synthesized idea brief from Stage 0}
Frameworks reference: {path to references/analysis-frameworks.md} — use PESTLE and TAM/SAM/SOM

OUTPUT:
Write report to $PROJECT_DIR/logs/product-ideation-market-research-{YYYYMMDD-HHMMSS}.md
Return summary to orchestrator: market size estimate, growth trend, top 3 key players, report path

Spawn: Task(subagent_type="product-ideation-market-researcher", prompt=...)

After both agents return: verify both log files exist. If either is missing, re-spawn that agent once.

Stage 2: Sequential — Competitive Analyzer

Read the market research log file path from Stage 1 output before constructing this prompt.

GOAL: Perform competitive intelligence analysis. Identify direct competitors, indirect
competitors, and failed competitors. Assess their positioning, pricing, strengths,
weaknesses, and the reasons for failure in failed cases.

CONSTRAINTS:
- Read market research output from: {market_research_log_path}
- Use WebSearch to discover competitors not mentioned in market research
- Include at least 3 direct competitors and 2 failed competitors
- Write output to $PROJECT_DIR/logs/product-ideation-competitive-analysis-{YYYYMMDD-HHMMSS}.md

CONTEXT:
Idea brief: {synthesized idea brief from Stage 0}
Market research: {market_research_log_path}
Frameworks reference: {path to references/analysis-frameworks.md} — use Porter's Five Forces

OUTPUT:
Write report to $PROJECT_DIR/logs/product-ideation-competitive-analysis-{YYYYMMDD-HHMMSS}.md
Return summary to orchestrator: top 3 competitors, market gap identified (yes/no), report path

Spawn: Task(subagent_type="product-ideation-competitive-analyzer", prompt=...)

Stage 3: Parallel Batch 2 — Pattern Documenter + Segment Analyzer

Read the competitive analysis log file path from Stage 2 output before spawning. Spawn both agents in the same message.

Pattern Documenter prompt structure:

GOAL: Document success patterns and failure patterns from the competitive landscape.
Produce competitor profiles, identify opportunity gaps, and extract repeatable lessons.

CONSTRAINTS:
- Read competitive analysis from: {competitive_analysis_log_path}
- May use WebSearch to deepen understanding of specific competitor trajectories
- Document at least 3 success patterns and 3 failure patterns
- Write output to $PROJECT_DIR/logs/product-ideation-patterns-{YYYYMMDD-HHMMSS}.md

CONTEXT:
Idea brief: {synthesized idea brief from Stage 0}
Competitive analysis: {competitive_analysis_log_path}
Frameworks reference: {path to references/analysis-frameworks.md}

OUTPUT:
Write report to $PROJECT_DIR/logs/product-ideation-patterns-{YYYYMMDD-HHMMSS}.md
Return summary to orchestrator: top success pattern, top failure risk, biggest opportunity gap, report path

Spawn: Task(subagent_type="product-ideation-pattern-documenter", prompt=...)

Segment Analyzer prompt structure:

GOAL: Identify and profile target user segments for this product idea. Determine who
will buy it, why, at what price, and in what volume. Produce TAM/SAM/SOM estimates
and at least 2 detailed user personas.

CONSTRAINTS:
- Read competitive analysis from: {competitive_analysis_log_path}
- Use WebSearch to validate segment size claims
- Produce at least 2 distinct user personas with willingness-to-pay estimates
- Write output to $PROJECT_DIR/logs/product-ideation-segments-{YYYYMMDD-HHMMSS}.md

CONTEXT:
Idea brief: {synthesized idea brief from Stage 0}
Competitive analysis: {competitive_analysis_log_path}
Frameworks reference: {path to references/analysis-frameworks.md} — use Jobs-to-be-Done + TAM/SAM/SOM

OUTPUT:
Write report to $PROJECT_DIR/logs/product-ideation-segments-{YYYYMMDD-HHMMSS}.md
Return summary to orchestrator: primary segment, estimated TAM, top persona name, report path

Spawn: Task(subagent_type="product-ideation-segment-analyzer", prompt=...)

After both agents return: verify both log files exist.

Stage 4: Sequential — Strategist

Read all four prior log file paths before constructing this prompt.

GOAL: Synthesize all research outputs into a definitive BUY/HOLD/SELL recommendation.
Produce the final report using the provided template. Support the recommendation with
confidence level, strategic rationale, risk factors, and next steps.

CONSTRAINTS:
- Read ALL of the following before writing:
    {idea_validation_log_path}
    {market_research_log_path}
    {competitive_analysis_log_path}
    {patterns_log_path}
    {segments_log_path}
- Use the final-report template at: {path to templates/final-report.md}
- Recommendation MUST be exactly one of: BUY, HOLD, or SELL
- Confidence MUST be exactly one of: High, Medium, or Low
- Write output to $PROJECT_DIR/logs/product-ideation-strategy-{YYYYMMDD-HHMMSS}.md

CONTEXT:
Idea brief: {synthesized idea brief from Stage 0}
All prior outputs: see CONSTRAINTS for paths
BUY/HOLD/SELL criteria: {path to references/analysis-frameworks.md} — see Recommendation Thresholds section
Final report template: {path to templates/final-report.md}

OUTPUT:
Write final report to $PROJECT_DIR/logs/product-ideation-strategy-{YYYYMMDD-HHMMSS}.md
Return summary to orchestrator: recommendation, confidence level, one-sentence rationale, report path

Spawn: Task(subagent_type="product-ideation-strategist", prompt=...)

After agent returns: read the full strategy report from the log path. Present its contents to the user verbatim (do not summarize — this is the deliverable).

Stage 5: Diagnostics (REQUIRED)

MANDATORY: Write diagnostic YAML after every invocation. Cannot be skipped.

Write to: $PROJECT_DIR/logs/diagnostics/product-ideation-{YYYYMMDD-HHMMSS}.yaml
Use schema from: templates/diagnostic-output.yaml

Include:
- Input type and value
- Interview rounds conducted
- Each pipeline stage: success | failed | skipped
- Final recommendation and confidence
- All log file paths
- Outcome: success | partial | failure

Error Handling

| Scenario | Action | |----------|--------| | Agent returns empty output | Re-spawn once with reinforced instructions. If still empty, mark stage as failed in diagnostics, skip to next stage. | | Idea validation returns FAIL | Present the FAIL verdict to user. AskUserQuestion: "The idea validator flagged significant concerns. Proceed anyway, or refine the idea?" | | Market researcher finds no data | Note data gap in final report. Strategist must lower confidence to Low if market size cannot be estimated. | | Both Stage 1 agents fail | STOP. Inform user: "Research pipeline failed at Stage 1. Please retry with a more specific idea description." | | Log file missing after agent completes | Glob for similar filenames in $PROJECT_DIR/logs/. If found, use it. If not, re-spawn agent. | | Token budget >60% at Stage 3 | Warn user. Spawn agents sequentially instead of in parallel to conserve context. | | Token budget >80% at any point | Stop at current stage. Present partial output with explanation. Write diagnostics noting partial completion. |

Token Budget Management

| Checkpoint | Threshold | Action | |------------|-----------|--------| | After Stage 0 interview | >20% consumed | Warn: "Significant context used in interview. Pipeline will consume more." | | After Stage 1 (parallel) | >45% consumed | Warn: "Approaching mid-budget. Consider running Stage 3 sequentially." | | After Stage 2 | >60% consumed | Warn: "Context budget constrained. Spawning Stage 3 sequentially." | | After Stage 3 | >75% consumed | Warn: "Low budget. Strategist prompt will be minimal." |

Completion Checklist

IMPORTANT: Before returning to user, verify ALL items are complete:

  • [ ] Stage 0: Arguments parsed (description or --doc)
  • [ ] Stage 0: references/analysis-frameworks.md loaded
  • [ ] Stage 0: templates/final-report.md loaded
  • [ ] Stage 0: AskUserQuestion interview conducted (1-2 rounds)
  • [ ] Stage 0: Synthesized idea brief produced
  • [ ] Stage 1: idea-validator spawned and output verified in $PROJECT_DIR/logs/
  • [ ] Stage 1: market-researcher spawned and output verified in $PROJECT_DIR/logs/
  • [ ] Stage 2: competitive-analyzer spawned and output verified in $PROJECT_DIR/logs/
  • [ ] Stage 3: pattern-documenter spawned and output verified in $PROJECT_DIR/logs/
  • [ ] Stage 3: segment-analyzer spawned and output verified in $PROJECT_DIR/logs/
  • [ ] Stage 4: strategist spawned and output verified in $PROJECT_DIR/logs/
  • [ ] Stage 4: Final report contents read from $PROJECT_DIR/logs/ and presented to user
  • [ ] Stage 5: Diagnostic YAML written to $PROJECT_DIR/logs/diagnostics/