Decompose complex implementation plans into focused, autonomous agent sprints. Each sprint is self-contained, has measurable acceptance criteria, and can execute without human intervention.
| >15 discrete steps or >8K tokens | Decompose into sprints | | Cross-domain work (multiple categories, services, or layers) | Decompose into sprints | | Long-running work requiring checkpoints | Decompose into sprints | | <10 steps, single domain, clear scope | Execute directly — decomposition adds overhead |
| Exploratory or research work | Execute directly — fluid reasoning needs cross-domain context |
Convert complex implementation plans into focused, autonomous agent sprints with measurable acceptance criteria. Use when (1) decomposing a large plan into agent-executable chunks, (2) breaking down multi-endpoint migrations or multi-step projects, (3) creating sprint definitions with acceptance criteria for autonomous agents, (4) planning dependency chains across implementation sprints, or (5) structuring work for parallel agent execution. Source: jkappers/agent-skills.