planning-and-task-breakdown

Decomposes specs into ordered, testable tasks with acceptance criteria and dependency-aware execution waves.

Updated Jun 13, 2026
One-click install
npx skills add https://github.com/22Teikk/22Teikk-Agent-Skills-Hub --skill planning-and-task-breakdown-22teikk
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: planning-and-task-breakdown
Source: https://github.com/22Teikk/22Teikk-Agent-Skills-Hub/tree/main/core/skills/planning-and-task-breakdown
Command: npx skills add https://github.com/22Teikk/22Teikk-Agent-Skills-Hub --skill planning-and-task-breakdown-22teikk

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Turning a specification into working code often fails because work is not broken into small, verifiable units. This Skill converts a spec or clear requirements into an ordered implementation plan where every task has acceptance criteria mapped to behavioral tests, explicit dependencies, and verification steps. ## Core Features & Use Cases - Vertical task slicing: Splits work into complete feature slices (data + domain + UI + tests) instead of horizontal layers, with a mandatory Foundation phase for Android projects (Hilt, Timber, Crashlytics). - AC-to-test traceability: Every acceptance criterion names the behavioral test that proves it, with a mandatory quality checklist covering empty input, failure, concurrency, and retry paths. - Parallel execution planning: Marks each task Parallel-safe and groups independent tasks into waves of up to four for concurrent execution via /teikk-build ultra. - Resumable artifacts: Writes plan.md plus a machine-readable todo.md task index for O(1) session resume, and a PARKING-LOT.md for deferred work that is re-surfaced each planning cycle. - Use Case: Given a spec for a task-management Android app, produce a phased plan where Task 1 is user registration (schema + API + UI + Room in-memory tests), with checkpoints after every 2-3 tasks. ## Quick Start Ask the agent to read the current spec and break the feature into an ordered implementation plan with acceptance criteria, verification steps, and a task index saved under .teikk/tasks/.

Frequently Asked Questions about planning-and-task-breakdown

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I break a feature spec into implementable tasks?▼

Read the spec in read-only plan mode, map the dependency graph, then slice vertically so each task delivers a working feature path (data, domain, UI, tests). Each task gets acceptance criteria mapped to behavioral tests, verification steps, dependencies, and a size estimate.

How to plan parallel task execution across multiple agents?▼

Mark each task's Parallel-safe field by checking that its dependencies are complete and its files-touched list does not overlap sibling tasks. Group independent tasks into waves of at most four; Foundation phase tasks are never parallel-safe.

What is vertical slicing vs horizontal slicing in task planning?▼

Vertical slicing builds one complete feature path at a time (schema plus API plus UI plus tests), so every task leaves the system working. Horizontal slicing builds all of one layer first, which delays testable functionality and hides integration problems.

When should I not use a formal task breakdown process?▼

Skip it for single-file changes with obvious scope, or when the spec already contains well-defined tasks. The planning overhead only pays off when work is large, vague, parallelizable, or spans multiple sessions.

How large should a single implementation task be?▼

Aim for Small (1-2 files) or Medium (3-5 files) tasks that fit one focused session. Anything Large (5-8 files) or bigger should be broken down further; a task title containing 'and' is a sign it is two tasks.

Why does a plan need a separate todo.md task index?▼

Long features outlive a single context window, so todo.md gives O(1) resume lookup: one line per task with checkbox state and a Current task pointer. It avoids re-reading the full plan.md every time a session restarts.