mark393295827/Atomic Decomposition

Atomic Decomposition Skill

Break complex problems into atomic, AI-solvable units with dependency tracking.

MECE+ Principles

| Principle | Description | |-----------|-------------| | Mutually Exclusive | No overlapping responsibilities | | Collectively Exhaustive | Complete coverage of the problem | | Atomic | Indivisible units of work | | Independent | Solvable without coordination |

Decomposition Template

problem: "[Problem statement]"

atomic_tasks:
  - id: T1
    action: "[Verb] [Object]"
    input: "[What's needed]"
    output: "[What's produced]"
    agent: "[Assigned agent]"
    deps: []
    
  - id: T2
    deps: [T1]
    # ...

Validation Checklist

Each atomic task must satisfy:

• [ ] Single responsibility (one action verb)
• [ ] Clear input/output contract
• [ ] No implicit dependencies
• [ ] Testable completion criteria
• [ ] < 100 word description

Dependency Graph Generation

graph LR
    subgraph "Parallel Group 1"
        T1[Task 1]
        T2[Task 2]
    end
    subgraph "Sequential"
        T3[Task 3]
    end
    T1 --> T3
    T2 --> T3

Common Patterns

Feature Implementation

Analyze → Design → Implement → Test → Document

Bug Fix

Reproduce → Diagnose → Fix → Verify → Regress

Refactoring

Audit → Plan → Extract → Transform → Validate

Usage

Called by manager-agent during objective decomposition:

skill: atomic-decomposition
input:
  objective: "[User goal]"
  constraints: "[Limitations]"
output:
  tasks: [atomic task list]
  graph: [dependency mermaid]