decompose
将问题分解为原子的、可测试的子问题。当面临复杂的任务、不明确的要求或规划实施策略时使用。
📖 中文导读:将问题分解为原子的、可测试的子问题。当面临复杂的任务、不明确的要求或规划实施策略时使用。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: decompose description: Break a problem into atomic, testable subproblems. Use when facing complex tasks, unclear requirements, or planning implementation strategy. allowed-tools: ["read_file", "update_plan"] metadata: short-description: Decompose problems systematically
Problem Decomposition
Read these references:
references/decomposition.md- Problem breakdown strategies~/.config/agent/core/task-management.md- Creating TODO lists from decomposition~/.config/agent/core/methodology.md- Integration with problem-solving framework
Instructions
Decompose problems into atomic, testable subproblems per references:
Decomposition Steps
Understand the whole problem
- Read all requirements and constraints
- Identify inputs, outputs, and transformations
- Note dependencies and assumptions
Identify natural boundaries
- Where can the problem be split?
- What are the logical units?
- Which parts are independent?
Break into subproblems
- Each subproblem should be:
- Atomic: Single, clear responsibility
- Testable: Can be verified independently
- Self-contained: Minimal external dependencies
- Name each subproblem clearly
- Each subproblem should be:
Map dependencies
- Which subproblems depend on others?
- What's the dependency graph?
- Which must be done sequentially?
Identify parallelization opportunities
- Which subproblems are independent?
- What can run concurrently?
- How can we maximize throughput?
Create TODO List
Per task-management.md, create a TODO list with:
- Each subproblem as an atomic task
- Dependencies clearly noted
- Parallelization opportunities marked
- Execution order specified
Example Output
TODO List for Problem X:
1. [pending] Subproblem A (no dependencies)
2. [pending] Subproblem B (no dependencies) [can parallelize with A]
3. [pending] Subproblem C (depends on A, B)
4. [pending] Subproblem D (depends on C)
Arguments
Problem to decompose: ${ARGUMENTS}