# Decompose Work

> Break approved product scope and architecture artifacts into epics, user stories, implementation tasks, dependency graph, critical path, ownership map, validation packs, merge order, and parallelization plan. Use after requirements and architecture are approved and before assigning implementation agents, opening parallel work, or writing artifacts `24-epics.md` through `29-agent-role-matrix.md`.

- Skill: `ashermahonin/decompose-work` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add ashermahonin/decompose-work`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ashermahonin/decompose-work/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: ashermahonin (https://skillmd.com/u/ashermahonin)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/ashermahonin/decompose-work

---


# Decompose Work

## Purpose

Make implementation boring in the best way. Convert a big plan into small, owned, sequenced tasks that can run without stepping on each other.

## Inputs

- Confirm requirements and architecture are approved enough to decompose.
- List deliverables and affected modules.
- Identify dependencies, shared files, contracts, and migration order.
- Decide which tasks can run in parallel and which must stay serial.

## Decision process

1. Create epics for coherent slices of value or infrastructure.
2. Break epics into stories and tasks with clear done criteria.
3. Assign each task a write scope, interfaces touched, validation command, and expected docs update.
4. Build the dependency graph and critical path.
5. Define parallel lanes only when file ownership and contracts are disjoint.
6. Write merge order, conflict policy, and fallback if a lane blocks.

## Decision boundaries

- Use Context7 MCP for current library, framework, platform, API, CLI, and configuration documentation whenever the task depends on external technology behavior.

## Decision record

- Epics
- Stories
- Tasks
- Dependency graph
- Parallelization plan
- Agent role matrix

## Ready when

- A task without validation is not ready.
- A parallel task without ownership boundaries is not parallel-ready.
- Do not split work by agent count; split by coherent boundaries.
- Keep merge order visible.

## Handoff

Hand off each implementation task with scope, files, contracts, validation, docs, dependencies, and merge order.

## References

- `references/parallelization-policy.md`: Use this before running multiple agents or parallel implementation lanes.