Guide · Agent skill
How to use the create-epics agent skill for game planning
Updated · written with AI from the project's own documentation and checked against our catalogue data
create-epics is an agent skill from the Donchitos Claude-Code-Game-Studios project that turns your game design documents and architecture notes into epics, one per architectural module. This guide is for solo and small indie developers, and for producers, who already have design docs and want a structured work plan. By the end you will know how to install the skill, trigger it with a plain request, approve or skip each draft epic, and what the skill will and will not touch.
Create Game EpicsDonchitos/Claude-Code-Game-Studios · MIT Download the skill Quick start
- 1Download and unzip. Download the ZIP from this page and unzip it so the folder is named create-epics and contains SKILL.md.
- 2Place the skill. Put the folder in ~/.claude/skills/ for all projects, or in your project's .claude/skills/ for one project. Or upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
- 3Confirm your project inputs. Make sure design/gdd/ holds your system GDDs and docs/architecture/architecture.md exists. Without them the skill stops and tells you what to create.
- 4Ask for epics. Make a plain request such as 'break my game into epics', or name a layer or system.
- 5Review each draft. Read each epic draft: its layer, module, governing ADRs and untraced requirements.
- 6Approve or skip. Answer yes, skip or pause for each epic. Only approved epics are written to production/epics/.
- 7Move on to stories. Run the story-creation skill on each epic, using its slug, to break it into stories.
What is an agent skill and what does create-epics do?
An agent skill is a folder with a SKILL.md file that gives an AI coding agent a specific task, rules and, sometimes, helper scripts. The agent reads the skill when your request matches its description.
create-epics reads your existing game design documents (GDDs) and architecture notes, then drafts one epic per architectural module. Epics are grouped by layer: Foundation, Core, Feature and Presentation. An epic describes what gets built for one module and which architecture decisions govern it. It does not describe how to build it; that is the job of stories.
- For each epic it shows you a draft before anything is saved.
- It writes an EPIC.md file and an index row only after you approve that epic.
- It checks which design requirements are backed by accepted architecture decisions and flags any that are not.
- It does not split an existing epic into stories. Use the story-creation skill for that.
Install the create-epics skill
Download the ZIP for this skill from this page and unzip it. You can install it in one of two places. For all your projects, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. The skill folder should be named create-epics and contain SKILL.md.
If you use the Claude apps instead, open Settings, go to Capabilities, then Skills, and upload the ZIP there.
- Personal install: ~/.claude/skills/create-epics/
- Project install: <your project>/.claude/skills/create-epics/
- Claude apps: Settings → Capabilities → Skills → upload the ZIP
Check that your project has what the skill needs
The skill is built for Claude Code or a compatible agent that supports SKILL.md skills. It needs a Bash shell because it calls helper scripts: gdd-structure-check.sh, adr-status.sh, md-section.sh, yaml-helper.sh and write-guard.sh.
It also expects the Claude Code Game Studios project layout. The inputs it reads are design/gdd/ for system GDDs, docs/architecture/architecture.md, and the .claude/scripts and .claude/docs folders. The source does not mention any API keys or external accounts.
- Claude Code or a compatible SKILL.md-capable agent
- A Bash shell
- design/gdd/ with system GDDs
- docs/architecture/architecture.md
- The .claude/scripts and .claude/docs folders
Trigger the skill with a natural request
You do not need to memorise a command. Ask in plain words when you want to plan work. The skill's description matches requests to break the game into epics or to plan the work before writing stories.
The skill also accepts an optional argument: all for every system, layer: foundation (or core, feature, presentation) for one layer, or a system name. Its argument hint also lists an optional --review setting with full, lean or solo. Your request can name a layer or system in words, such as 'foundation and core'.
- "Break my game into epics."
- "Plan the work for the Core layer before I write stories."
- "Create epics for the combat system."
- "Turn my GDDs and architecture into epics, one per module."
Review and approve each epic
The skill works through layers in order: Foundation, then Core, Feature and Presentation. For each system it shows a draft with its layer, GDD, architecture module, governing ADRs, engine risk and how many requirements are covered by accepted decisions.
It then asks, for that one epic only, whether to create it. Your options are to create it, skip it, or pause because you need to write architecture decisions first. Nothing is written until you say yes. Each approved epic is saved to production/epics/[epic-slug]/EPIC.md and added to production/epics/index.md.
- Yes, create it: writes that epic's EPIC.md and index row
- Skip this epic: moves on without writing it
- Pause: lets you write architecture decisions first
Understand requirement coverage and warnings
Each requirement (TR-ID) for a system is either covered or untraced. A requirement counts as covered only when an Accepted architecture decision addresses it. If the only decision is still Proposed, the requirement is untraced, so the story step will still see that decision and block its story.
When requirements are untraced, the skill lists them and may show a warning, depending on the rigour level of your project. Your options are to run the architecture-decision skill first or accept the gap.
- Covered: an Accepted ADR addresses the requirement
- Untraced: no Accepted ADR addresses it, including Proposed ones
- Warnings at the strictest level apply to every untraced requirement; at the lightest, to Foundation-layer ones
Safety notes: scripts and network
This skill runs scripts. The source lists helper scripts that the skill may call through Bash, including gdd-structure-check.sh, adr-status.sh, md-section.sh and write-guard.sh. Review them in the skill folder before you install if you need to know exactly what they do.
The source says the skill does not use the network. It writes files only after you approve each one, and it does not write stories.
- Runs scripts: yes
- Uses network: no
- Writes only after your yes for each epic
What people use it for
- A solo developer turning finished design docs into a list of module epics before starting code.
- A small team on the Claude Code Game Studios workflow moving from design into production, one layer at a time.
- A producer checking that each design requirement traces to an accepted architecture decision before stories are written.
- Re-running the skill for the Core layer once Foundation decisions are accepted.
Frequently asked questions
Is create-epics free to use?
The source file is released under the MIT licence. The source does not mention any price or paid tier for this skill.
What do I need before I can use create-epics?
You need Claude Code or a compatible SKILL.md agent, a Bash shell, system GDDs in design/gdd/, and docs/architecture/architecture.md. If the design or architecture documents are missing, the skill stops and tells you to create them first.
Where do I install the create-epics skill?
Unzip the downloaded ZIP into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder for one project. In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
Does create-epics write stories?
No. It writes an EPIC.md per approved epic and its index row. Stories are created afterwards with the story-creation skill, such as /create-stories with the epic's slug.
Why does the skill say a requirement is untraced?
A requirement is untraced when no Accepted architecture decision covers it. If the only decision is Proposed, it still counts as untraced, so the story step will block that story until the decision is accepted.
Can I use create-epics for a non-game project or to split an existing epic?
Not for either. Its inputs and terms are built around game design documents, and it does not split epics into stories. Use the story-creation skill to break an epic into stories.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use create-epics if you already have game design documents and an architecture document and want a structured, traceable work plan before writing stories. Pick something else if you are at the very start with no design or architecture docs, if you need to split an epic into stories, or if your project is not a game.
Download the skill