Planning and Task Breakdown guides an agent from a spec to a task list it can execute. The agent first works in a read-only plan mode, maps the dependency graph, slices the work vertically and writes each task with acceptance criteria, verification steps, dependencies and likely files.
Tasks are ordered so risk comes early, with checkpoints after every two or three tasks. The plan goes to tasks/plan.md and the task list to tasks/todo.md or a designated external tracker, and an incomplete plan is never overwritten without asking.
What You Get
- A five-step planning process
- A task template and checkpoint template
- A plan document template with risks and open questions
- Parallelization guidance and a verification checklist
Common Use Cases
- Breaking a spec into implementable tasks
- Splitting a task that feels too large
- Ordering work by dependency and risk
- Recording tasks in an external tracker
What Makes It Unique
The skill protects existing plans: it checks for unchecked tasks before writing and stops to ask when the new work differs. It also maps the same task structure onto an external tracker and keeps plan.md as an index of tracker items.
Under The Hood
The skill is a single SKILL.md of instructions. The agent reads the spec and codebase in a read-only phase, then writes tasks/plan.md and tasks/todo.md, creating the tasks folder if needed. With a designated tracker it creates items there. It runs no scripts.