mav-create-tasks

Create Tasks

Decompose a solution design into discrete, independently implementable tasks. Each task is small enough to be implemented, verified, and committed in a single focused effort.

Principles

1. Each task is independently implementable — it can be completed without waiting for other tasks (unless an explicit dependency exists)
2. Each task is independently committable — it leaves the codebase in a working state
3. Each task is small and discrete — a title and 1-2 sentence description is enough to act on
4. Tasks are ordered by dependency — later tasks build on earlier ones
5. The sum of all tasks satisfies the design — nothing is missing, nothing is extra

Process

digraph tasks {
    "Read the solution design" [shape=box];
    "Identify discrete units of work" [shape=box];
    "Order by dependency" [shape=box];
    "Count tasks" [shape=diamond];
    "Post checklist comment" [shape=box];
    "Create sub-issues" [shape=box];

    "Read the solution design" -> "Identify discrete units of work";
    "Identify discrete units of work" -> "Order by dependency";
    "Order by dependency" -> "Count tasks";
    "Count tasks" -> "Post checklist comment" [label="< 5 tasks"];
    "Count tasks" -> "Create sub-issues" [label=">= 5 tasks"];
}

1. Identify Discrete Units of Work

From the solution design's "Areas Affected" and "Approach", identify natural task boundaries:

• A new file or module
• A modification to an existing function or component
• A new test suite or test additions
• A configuration or infrastructure change
• A database migration

Each task should touch a small, cohesive set of files. If a task feels like it needs a paragraph to describe, it is too big — split it.

2. Order by Dependency

Arrange tasks so each builds on the previous:

• Data models before business logic
• Business logic before API endpoints
• API endpoints before UI integration
• Implementation before tests (or tests first if using TDD)