AgentSkillsCN

Role Decomposition

提供一套方法论与决策框架,用于将团队职责细分为不同智能体角色,并合理把控分解的粒度。

SKILL.md
--- frontmatter
name: Role Decomposition
description: Provide methodology and decision framework for decomposing team responsibilities into agent roles with appropriate granularity

Role Decomposition

Description

Provide methodology and decision framework for decomposing team responsibilities into agent roles with appropriate granularity.

Users

  • role-designer: For responsibility decomposition during role design phase
  • team-architect: For structural rationality assessment during review phase

Core Framework: MECE Responsibility Decomposition

Each agent's responsibilities must follow the MECE principle:

  • Mutually Exclusive: Any specific work should belong to only one agent
  • Collectively Exhaustive: All work combined should have no omissions

Granularity Guidance

For detailed quantitative granularity assessment, use skills/granularity-calibration/SKILL.md. Quick reference:

  • Too coarse → Split: Agent has 5+ core work items, spans 2+ professional domains, or cannot be described in one sentence
  • Too fine → Merge: Two agents always communicate on every task, or one agent has only 1-2 simple tasks
  • Appropriate: Each agent has one clear sentence describing their core responsibility, 3-5 work items, and interactions are primarily "deliverable handoffs"

Grouping Strategy

Choose grouping criteria by priority:

  1. Workflow stages (Most recommended)

    • Applicable scenario: Team work has clear sequential order
    • Example: discovery → planning → execution → review
  2. Professional domains

    • Applicable scenario: Team spans multiple specialties, but processes within each are similar
    • Example: content → technical → quality
  3. Deliverable-oriented

    • Applicable scenario: Team's deliverable types are significantly different
    • Example: code → documentation → visual-assets