kvasir-task-decomposition

Break complex objectives into well-scoped subtasks with explicit dependencies and execution order.

SKILL.md

GitHub

Task Decomposition

Purpose

Transform complex objectives into well-defined, independently executable subtasks with clear dependencies, inputs, and outputs. Ensure each subtask maps to exactly one owner with one well-defined output.

When to Use

  • When a task spans multiple workstreams or requires multi-step execution.
  • When dependencies between subtasks are non-trivial.
  • When the execution order matters and blocking relationships exist.
  • When a task is large enough that decomposition reduces risk and improves clarity.

Workflow

  1. Analyze the objective.

    • Clarify the end goal and what a successful outcome looks like.
    • Identify constraints such as time, tooling, and architecture.
    • Note which capabilities the requesting agent has made available.
  2. Identify atomic subtasks.

    • Break the objective into the smallest meaningful units of work.
    • Ensure each unit produces a single, well-defined output.
    • Confirm no subtask bundles unrelated concerns together.
  3. Map dependencies.

    • Determine which subtasks depend on the outputs of others.
    • Distinguish hard blocking relationships from soft ordering preferences.
    • Identify work that can proceed in parallel.
  4. Assign ownership.

    • Match each subtask to the required capability based on task type.
    • Verify the required capabilities are available.
    • Ensure every subtask has exactly one responsible owner.
  5. Define execution order.

    • Sequence subtasks so dependencies are satisfied before dependents run.
    • Group parallelizable subtasks to shorten the critical path.
    • Highlight blocking points where downstream work must wait.
  6. Document the plan.

    • Assign each subtask a stable ID and a clear description.
    • Record ownership assignments, dependencies, and expected outputs.
    • Present the decomposition so it can be executed without ambiguity.

Quality Criteria

  • Every subtask has a single output and a single responsible owner.
  • All dependencies are explicitly stated, not implied.
  • Parallelizable subtasks are identified.
  • The plan covers the full objective with no gaps.
  • Each subtask is small enough to be unambiguous.

Anti-Patterns

  • Mixed outputs: A subtask that bundles research and implementation into one unit.
  • Implicit dependencies: Assuming a subtask can start without checking its inputs are ready.
  • Over-decomposition: Splitting work so finely that coordination overhead exceeds the work itself.
  • Missing ownership: A subtask without a clear owner or required capability.