skills.rest
CagesThrottleUs
CagesThrottleUsCOMMUNITY
Explore all skills by this author →
GitHub

epic-decomposition

Breaks an approved epic into INVEST-compliant stories with dependency waves and child manifests.

Updated May 21, 2026
Search Tags:#user stories#agile planning#invest#dependency graph#epic decomposition#hld#spidr
One-click install
npx skills add https://github.com/CagesThrottleUs/private-ai-harness --skill epic-decomposition-cagesthrottleus
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: epic-decomposition
Source: https://github.com/CagesThrottleUs/private-ai-harness/tree/main/skills/epic-decomposition
Command: npx skills add https://github.com/CagesThrottleUs/private-ai-harness --skill epic-decomposition-cagesthrottleus

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Large epics are too big to implement in one pass, and ad-hoc splitting produces horizontal layer stories that cannot ship independently. This Skill derives bounded, vertically sliced stories directly from an approved HLD so each story can run through the task lane on its own. ## Core Features & Use Cases - HLD-driven decomposition: Maps each C4 container in the approved HLD to a story, enforcing INVEST and vertical-slice rules with SPIDR splitting patterns. - Dependency mapping: Labels story pairs as blocks or informs, then groups stories into parallel execution waves. - Child manifest generation: Creates a manifest per story recording tier, gate status, dependencies, produced interfaces, spec scope, and HLD containers. - Use Case: After a human approves the HLD for a notifications epic, run this Skill to produce stories like "user gets an email on order-ship", a wave plan, and child manifests, then hand each story to the task lane. ## Quick Start Ask the AI to decompose the approved epic into stories using the epic-decomposition skill after the HLD has been human-approved.

Frequently Asked Questions about epic-decomposition

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

FAQPage Schema
How do I break an epic into user stories?▼

Derive stories from the approved HLD's C4 Container diagram, mapping each container to one vertical slice that passes INVEST criteria. Split oversized capabilities using SPIDR patterns: Spike, Paths, Interfaces, Data, and Rules.

What is a vertical slice story versus a horizontal layer story?▼

A vertical slice delivers an observable end-to-end behavior change, potentially touching DB, API, and UI together. Horizontal slices like "the API story" fail INVEST because none is independently valuable and they force lockstep integration.

Can epic decomposition run before the HLD is approved?▼

No. The HLD C4 Container diagram is the decomposition map, so the skill hard-gates on human HLD approval. Decomposing earlier produces invented stories that drift from the actual architecture.

How are parallel and sequential story dependencies determined?▼

Each story pair is labeled blocks or informs: blocks means the downstream story needs an upstream committed artifact, while informs allows parallel work if the contract is stable. Stories with no upstream blockers form Wave 1 and run in parallel.

What are the limitations of epic decomposition?▼

It requires an approved HLD, spec, and epic manifest as inputs and does not produce task-level plans, which come from writing-plans inside each story. Stories that depend on everything or lack a testable done-state must be split further.

Related Skills

Explore More Software Engineering Skills→
cyariecyarie

breaking-apart-systems

Decompose system descriptions into C4 architecture diagrams with gated approvals.

Community
Advanced
wellcrafted-devwellcrafted-dev
35

spec-execution

Execute specification documents through planned waves of parallel and sequential changes.

Official
Advanced
Rust-sohamRust-soham

golem-parallel-workers-ts

Distribute independent work across Golem agent instances and collect results.

Community
Intermediate
thesyntaxninjathesyntaxninja
2

ralph-loop

Decompose oversized tasks into verifiable pass/fail stories for iterative execution.

Community
Advanced