Skip to content
Back to skills

Mav Create Tasks

ASecurity

How to decompose a solution design into discrete, independently implementable tasks. Used as a dependency from workflow skills.

  • 11 stars
  • 0 votes
  • 0 copies
  • 4 views
  • Added June 1, 2026
documentationbashgitapidatabase

Works with

  • api

Security analysis

A100/100

Scanned June 1, 2026

npx -y skills add thermiteau/maverick --skill mav-create-tasks --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Mav Create Tasks?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Mav Create Tasks
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/thermiteau-mav-create-tasks/badge)](https://www.skillsdirectory.com/skills/thermiteau-mav-create-tasks)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: mav-create-tasks
description: How to decompose a solution design into discrete, independently implementable tasks. Used as a dependency from workflow skills.
---

# 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

```dot
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)

### 3. Choose Output Format

The output format depends on the number of tasks:

#### Less than 5 tasks: Checklist Comment

Post a single comment on the issue with a checkbox list:

```markdown
## Tasks

- [ ] **<imperative title>** — <1-2 sentence description>
- [ ] **<imperative title>** — <1-2 sentence description>
- [ ] **<imperative title>** — <1-2 sentence description>
```

Each task is checked off as it is completed. Progress is tracked by updating the comment.

#### 5 or more tasks: Sub-Issues

Create GitHub sub-issues linked to the parent issue. Each sub-issue contains:

- A clear, imperative title
- A 1-2 sentence description of what to implement
- A reference to the parent issue for context

```bash
gh issue create \
  --title "<imperative title>" \
  --body "$(cat <<'EOF'
<1-2 sentence description of what to implement>

Parent: #<parent-issue>
EOF
)"
```

Post a summary comment on the parent issue listing all sub-issues with their execution order:

```markdown
## Tasks

| # | Issue | Title | Depends On |
|---|-------|-------|------------|
| 1 | #101 | <title> | — |
| 2 | #102 | <title> | — |
| 3 | #103 | <title> | #101 |

**Execution order:** #101, #102 → #103
```

Sub-issues that have no dependency between them can be worked on in any order. Use the `Depends On` column only when one task genuinely requires another to be completed first.

## Task Sizing

### Right Size

- **1-3 files** touched per task
- **Single concern** — does one logical thing
- **Self-evident** — the title and a short description are enough to implement it without further context (the solution design provides the broader context)

### Too Large

- "Update the API layer" — which endpoints? Split by endpoint or concern.
- "Add tests" — for what? Split by component under test.
- "Refactor the module" — what specifically? Split by function or concern.

### Too Small

- "Add import statement" — combine with the task that uses the import.
- "Create empty file" — combine with the task that adds content.
- "Update a single config line" — combine with related changes.

## Validation

Before the task list is considered complete:

- [ ] Every acceptance criterion from the issue is addressed by at least one task
- [ ] No task is so vague that it requires a design of its own
- [ ] Dependencies between tasks are minimal and explicitly stated
- [ ] The sum of all tasks fully implements the solution design

<!-- maverick-plugin-version: 3.3.5 -->

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.