Task Objectives
- This Skill is used to: simulate multiple AI agents with different professional backgrounds and personality traits conducting meeting discussions, ultimately reaching a decision consensus
- Capabilities include:
- Customize meeting topics and goals
- Configure multiple agent roles (role name, professional field, personality traits)
- Simulate a real meeting discussion process (speeches, debates, supplements, summaries)
- Display the meeting speech process in real time
- Generate a complete meeting record including decision conclusions, key arguments, risk warnings, and follow-up actions
- Trigger conditions: when users need to analyze a problem from multiple perspectives, need collisions of professional knowledge from different fields, or need to simulate team decision-making discussions
Operating Steps
1. Determine the meeting topic and goals
Extract the following information from user input:
- Meeting topic: the core issue to be discussed (such as "whether to adopt a microservices architecture", "new product pricing strategy")
- Meeting goal: the type of decision expected to be reached (such as feasibility assessment, solution selection, risk identification)
- Key constraints: constraints that need to be considered (such as budget, time, technology stack, etc.)
2. Design agent roles
Based on the meeting topic, design 3-5 agent roles, each of which should include:
- Role name: such as "Technical Architect", "Product Manager", "Financial Advisor"
- Professional field: clarify their professional background and focus
- Personality traits: such as rational and rigorous, aggressively innovative, conservative and steady
- Speaking style: concise and direct or detailed argumentation
Refer to references/agent-roles.md role templates, and select or create a suitable role combination based on the meeting topic.
3. Execute the meeting process
Simulate the meeting discussion according to the following stages:
Stage 1: Opening speeches
- The moderator (agent) introduces the meeting topic and goals
- Each agent gives an opening statement from their own professional perspective
Stage 2: Free discussion
- Agents respond, question, and supplement based on others' speeches
- Encourage collisions and debates among different viewpoints
- Guide in-depth exploration of key issues
Stage 3: In-depth debate
- Conduct targeted debates on the focal points of controversy
- Cite professional arguments and data to support viewpoints
- Identify and discuss potential risks
Stage 4: Consensus convergence
- Summarize the common points and differences among all parties' viewpoints
- Explore compromise solutions and alternative solutions
- Guide convergence toward the decision direction
Stage 5: Decision generation
- Integrate opinions from all parties to form a decision conclusion
- Clarify the main basis for the decision
- Identify the risks and precautions of the decision
4. Output the meeting record
According to references/meeting-record-format.md Output meeting minutes in the format defined in
- Basic meeting information
- Speech process record (displayed in real time)
- Decision conclusions and key arguments
- Risk alerts and follow-up action recommendations
Resource index
- Role templates: see references/agent-roles.md(provides common role types and characteristic descriptions)
- Minutes format: see references/meeting-record-format.md(defines the standard output format for meeting minutes)
- Meeting templates: see assets/meeting-templates/(provides preset role configurations for different types of meetings)
Usage examples
Example 1: Technical architecture decision meeting
- Function description: Discuss whether to adopt a microservices architecture
- Execution method: Agent-led, no script required
- Role configuration: Technical architect, DevOps engineer, frontend engineer, CTO
- Output content: Detailed technical debate, cost analysis, risk assessment, final recommendation
Example 2: Product pricing strategy meeting
- Function description: Formulate a pricing strategy for a new product
- Execution method: Agent-led, no script required
- Role configuration: Product manager, market analyst, sales director, financial advisor
- Output content: Market competition analysis, pricing plan comparison, expected revenue assessment
Example 3: Project risk assessment meeting
- Function description: Identify and assess project risks
- Execution method: Agent-led, no script required
- Role configuration: Project manager, technical expert, legal advisor, operations expert
- Output content: Risk identification checklist, impact assessment, recommended response measures
Notes
- Ensure that each agent's speech matches its role setting and professional background
- Maintain the authenticity and logic of the discussion and avoid empty opinions
- Focus on the collision and debate process of different viewpoints, as this is the core value of the meeting
- Decisions should be based on sufficient arguments and consideration, reflecting comprehensive analysis from multiple perspectives
- Meeting minutes should be clear and easy to read, with emphasis on the decision-making process and conclusions
- Make full use of the agents' language understanding and generation capabilities, and do not introduce scripts for simple tasks