Task decomposition breaks a complex problem into smaller subtasks that can be handled separately. In agent systems, it can mean identifying the steps needed for a novel problem, not just following a preset template. section 5.0 section 10.1.3
Decomposition identifies what the pieces are; planning decides their order, dependencies, and approach. These operations can fail independently, so a poor result may come from a flawed breakdown or from a bad plan for sound subtasks. section 5.0
From everyday steps to agent subtasks
For “leaving home,” a simple breakdown might include getting dressed, checking the toaster, and grabbing keys; planning puts those steps in sequence. For a research goal, an orchestrator can delegate subtasks to workers, then add or reprioritize them if findings reveal gaps in the original breakdown. section 5.1 section 9.3
Table 10.1 Agent cognitive capabilities
| Capability |
Present |
Absent |
Supporting primitive |
| Task decomposition |
Agent invents novel subtask sequences |
Agent treats complex queries as flat lookups |
CoT, planning |
| Dependency reasoning |
Agent sequences steps while respecting dependencies |
Agent executes steps in arbitrary order |
ToT, sequential thinking |
| Compositional tool use |
Agent chains tools in novel configurations |
Agent maps queries to single tool calls |
ReAct |
| Model updating |
Agent revises plan when evidence contradicts it |
Agent follows initial plan regardless |
Reflexion, replanning |
Compare task decomposition with other cognitive capabilities, their observable effects, and their supporting reasoning primitives. section 10.1.3
For straightforward tasks whose breakdown is already known, hardcoded task dispatch is cheaper and faster than an orchestrator. Reserve dynamic orchestration for goals with uncertain subtasks or plans that may need to change as results arrive. section 9.3
