• What does the user want to analyze? (Review data over a period, a product strategy, a business direction, a decision, etc.)
• Are there any existing material files? (If yes, ask the user to provide the path or content)

• What does the user want to gain from the analysis? (Diagnose problems, verify directions, discover blind spots, make decisions, etc.)
• Are there any core questions the user specifically wants answered?

• Use all 10 default perspectives: Suitable for comprehensive diagnosis ("The world's 10 smartest minds act as your consultants simultaneously")
• Select 3-5 of them: Suitable for focusing on specific dimensions
• Customize perspectives: Users specify other thinkers/frameworks, and AI temporarily builds corpora via WebSearch

• Clear analysis object (with materials or clear description)
• Clear analysis purpose (at least one core question)
• Confirmed perspectives

1. If the user provides a file path: Read the file content and organize it into a text summary
2. If the user describes in dialogue: Extract key information to form a structured "Analysis Brief"
3. If the material is too long (over 5000 words): Extract core data and key facts, compress into a version digestible for each agent

1. Role Setting: You are [Thinker's Name], analyze using the methodology of [Framework Name]
2. Analysis Material: [Organized user material]
3. Core Question: The user's most desired question to answer
4. Framework Reference: Full original text of the corresponding reference file
5. Output Requirements:

• Do not pre-inject the full reference text into the main prompt — stuffing 5 corpora into the prompt will make it too long, causing agent loss or degradation
• Instead: In each agent's prompt, tell it to read the corresponding reference file on its own
• Each agent only reads its own file, keeping the workload manageable
• The reference file includes original quotes from the thinker, specific cases, and analysis question templates — the agent must conduct analysis based on these
  • Must use core concepts and terminology of the framework
  • Must provide clear judgments and suggestions, avoid ambiguity
  • Must include "The Harshest Truth" — a diagnosis the user may not want to hear but must listen to
  • No word limit; write as much as needed
  • Tone: Speak directly to the user in the first person of the thinker, like a real consultant

• The main dialogue does not pre-read reference files (to avoid prompt length causing agent loss)
• Each agent reads its own reference file independently
• After analysis, each agent writes its report to the specified file on its own
• The main agent only reads all reports at the end for cross-summary

.claude/skills/multi-perspective-analysis/reference/01-10x-growth-dan-sullivan.md

多视角分析/分析报告-[主题]/01-DanSullivan.md

.claude/skills/multi-perspective-analysis/reference/02-first-principles-elon-musk.md

多视角分析/分析报告-[主题]/02-马斯克.md

.claude/skills/multi-perspective-analysis/reference/03-product-intuition-zhang-xiaolong.md

多视角分析/分析报告-[主题]/03-张小龙.md

.claude/skills/multi-perspective-analysis/reference/04-creator-growth-mrbeast.md

多视角分析/分析报告-[主题]/04-MrBeast.md

.claude/skills/multi-perspective-analysis/reference/05-mental-models-charlie-munger.md

多视角分析/分析报告-[主题]/05-芒格.md

.claude/skills/multi-perspective-analysis/reference/06-zero-to-one-peter-thiel.md

多视角分析/分析报告-[主题]/06-蒂尔.md

.claude/skills/multi-perspective-analysis/reference/07-product-vision-steve-jobs.md

多视角分析/分析报告-[主题]/07-乔布斯.md

.claude/skills/multi-perspective-analysis/reference/08-customer-obsession-jeff-bezos.md

多视角分析/分析报告-[主题]/08-贝佐斯.md

.claude/skills/multi-perspective-analysis/reference/09-algorithm-thinking-zhang-yiming.md

多视角分析/分析报告-[主题]/09-张一鸣.md

.claude/skills/multi-perspective-analysis/reference/10-strategic-survival-ren-zhengfei.md

多视角分析/分析报告-[主题]/10-任正非.md

You are [Thinker's Name]. Use your [Framework Name] to analyze the following material and answer the user's core question.

1. Conduct in-depth analysis using the core concepts of your framework; avoid superficial discussions
2. Must quote original content from the reference file to support your analysis (marked with > quote format)
3. Provide clear judgments and suggestions
4. Speak directly to the user in the first person, like a real consultant
5. Must include a section of "The Harshest Truth" — a diagnosis that the user most needs to hear but may least want to listen to
6. Write as much as needed; prioritize depth over length control
7. Please output in Chinese

• Create folder first before launching: Only start agents after mkdir is completed
• Must launch all in parallel: All selected agents must be launched in the same message. If 5 perspectives are selected, 5 agents must be launched without omission
• Each agent writes its own file: The main agent is not responsible for writing perspective-specific reports, only the summary
• Each agent is completely independent: Do not let any agent see the output of other agents
• Do not preset conclusions: Do not imply "You should find XX problem" in the prompt

• Scan all reports to find points where 3 or more perspectives reach the same conclusion
• These are the most reliable diagnoses — convergence of independent analyses indicates "hard facts"

• Identify points where different perspectives provide contradictory suggestions
• Analyze the reasons for differences: Are they due to different framework assumptions? Or different levels of the problem being observed?
• Differences are often more enlightening than consensus — they expose the multi-faceted nature of the problem

• Did any perspective propose an important point that no one else mentioned?
• These may be overlooked blind spots

• Extract specific action recommendations from all reports
• Sort by "Importance x Executability"
• Converge into 3-5 most critical next steps

Analysis Date: YYYY-MM-DD Perspectives: [Perspectives used] Material: [What was analyzed]

多视角分析/

多视角分析/
  分析报告-[主题简称]/
    00-汇总报告.md
    01-Dan-Sullivan.md
    02-Elon-Musk.md
    03-Naval-Ravikant.md
    04-张小龙.md
    05-MrBeast.md

00-汇总报告.md

• Which directory the files are saved in
• Total number of reports
• Display the core content of the summary report in the dialogue (consensus + differences + action recommendations)

• Folder name:

  分析报告-[主题简称]

  (e.g.,

  分析报告-职业决策

  ,

  分析报告-4月策略

  )
• If a folder with the same topic already exists, add the date after the topic:

  分析报告-职业决策-0410

• If there are custom perspectives (corpora built via WebSearch), save them to the reference directory for reuse

1. Independence is the lifeline: Each agent must conduct completely independent analysis and cannot see the output of other agents. This is the entire value of multi-perspective analysis
2. Must be parallel: All agents must be launched in the same message, not serially. Serial execution will cause later agents to be influenced by earlier ones
3. Depth first: It is better to have a 3000-word in-depth report for each perspective than 5 shallow 500-word summaries
4. Must have sharpness: Each agent must provide "The Harshest Truth". Polite but useless analysis is better not done
5. Summary must have opinions: Cross-summary is not simple listing; analyze why consensus formed and why differences occurred
6. User decision-making power: Analysis is a tool, decision-making belongs to the user. Provide suggestions at the end of the report, do not make decisions