Guide · Agent skill
How to use the Create Stories agent skill (create-stories)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The create-stories skill is an agent skill for Claude Code (and compatible agents) that breaks one epic, a large feature chunk, into small implementable stories. Each story gets acceptance criteria, which are checkable conditions for done, plus a link back to the requirement and architecture decision that governs it. It is built for a game studio workflow based on planning documents, so this guide explains what it does, how to install it, how to trigger it, and what it needs before you rely on it.
Create StoriesDonchitos/Claude-Code-Game-Studios · MIT Download the skill Quick start
- 1Download the ZIP. Download the create-stories ZIP from this page.
- 2Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 3Check your folder structure. Confirm the project has production/epics, design/gdd and docs/architecture.
- 4Make sure an epic exists. Confirm the epic has an EPIC.md file and that its governing ADRs exist.
- 5Ask for stories by epic name. Ask the agent to break a named epic into stories, or pass the epic slug or path.
- 6Answer the questions. Answer each question the skill asks; it writes files only after you confirm.
- 7Run the next step. Follow the suggested next step, such as a readiness check and then development.
What an agent skill is and what create-stories makes the agent do
An agent skill is a folder with a SKILL.md file that tells the agent how to handle a specific kind of task. Once the skill is installed, the agent can follow its process when your request matches the skill's description.
The create-stories skill has one job: break a single epic into story files. Its description says it is for stories for an epic or the next batch of work to build, and it is not for turning a whole design into epics. Following its process, the agent reads the epic's design document, the relevant architecture decisions and rules, then writes small story files. It also updates the epic's story table and an index file, and suggests next steps such as readiness checks and development.
- Reads the epic's design document and the architecture decisions and rules that govern it
- Writes one story file per implementable behaviour, each with acceptance criteria
- Traces each story to a requirement and the decision that governs it
- Updates the epic's story table and an index file
- Suggests next steps such as readiness checks and development
Install the create-stories skill
Download the ZIP for the create-stories skill from this page. Then choose one of the two install locations below.
For a personal install available to all your projects, unzip the download into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. Both paths put the skill folder where Claude Code can find it.
If you use the Claude apps instead, open Settings, then Capabilities, then Skills, and upload the ZIP there.
- Personal: unzip into ~/.claude/skills/
- Single project: unzip into your project's .claude/skills/
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Trigger the skill with natural requests
You do not need a special command. Ask in plain words for stories for a named epic, or for the next batch of work to build. The skill also accepts an epic slug or a path as its argument, using the form [epic-slug | epic-path] with an optional --review full, lean or solo.
Because the skill is for breaking an epic down, ask for the epic by name. If you ask for the whole game design to be split into epics, the skill is the wrong tool; its own description says that is handled elsewhere.
- "Break the combat epic into stories."
- "Write the next batch of stories for the loot epic."
- "Create stories for production/epics/combat/EPIC.md."
What the skill needs before it runs
The skill requires Claude Code or a compatible agent with file read and write access, and a bash shell, because it runs helper scripts in .claude/scripts and .claude/hooks. It also expects the Claude Code Game Studios folder structure: production/epics, design/gdd and docs/architecture.
If those folders are missing, the skill will stop or fail to find its inputs. Its inputs include the epic's EPIC.md file, the design document under design/gdd, and the architecture decision records (ADRs) in docs/architecture.
- Claude Code or a compatible agent with file read/write access
- A bash shell
- production/epics, design/gdd and docs/architecture folders
- An epic file, design document and architecture decision records for the epic
Safety notes: scripts and network
The skill runs scripts: yes. Its helper scripts are the ones in .claude/scripts and .claude/hooks that it is allowed to call. Review those scripts in your project before running the skill on work you care about.
The skill does not use the network, according to its listing.
- Runs helper scripts: yes
- Uses network: no
- Review the scripts in .claude/scripts and .claude/hooks before first use
Tips for getting useful stories
Expect the skill to ask questions. Its process asks multiple questions and writes several files, so it is heavier than a quick task list. Answer them, since a skipped question is not treated as an answer.
Choose the story size with the story_granularity setting. Coarse stories hold 5 to 10 acceptance criteria and suit a whole feature. Balanced stories hold 2 to 4 and suit one task. Fine stories hold one criterion each. Run the skill once per epic, in dependency order: Foundation epics first, then Core, Feature and Presentation.
Use the QA lead review when you want stronger checks on testability. It runs only when the review gate is enabled in full mode.
- Run one epic at a time, Foundation first
- Pick a story size with story_granularity: coarse, balanced or fine
- Use --review to set the review mode for a run
- Expect several questions and several files written per run
What people use it for
- Splitting a combat epic into implementable stories with acceptance criteria
- Building the next batch of work from a loot or inventory epic
- Keeping acceptance criteria consistent across a backlog for a small team
- Tracing each story back to the design requirement and architecture decision that governs it
Frequently asked questions
Is the create-stories skill free?
This page does not state a price or licence terms beyond the listing's licence: MIT. Check the repository's licence file for the full terms before reuse.
What do I need to install create-stories?
You need Claude Code or a compatible agent with file read and write access, and a bash shell. The project should follow the Claude Code Game Studios folder structure with production/epics, design/gdd and docs/architecture.
Where do I install the create-stories skill?
Unzip the downloaded ZIP into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, upload the ZIP under Settings, then Capabilities, then Skills.
Can create-stories turn a whole game design into epics?
No. The skill says it breaks one epic into stories and is not for turning the whole design into epics. Use the epic-creation skill for that step.
Why does create-stories stop or fail?
The most likely cause is a missing input. The skill needs the epic's EPIC.md, the design document and the architecture decision records; if your project does not use the expected production/epics and design/gdd folders, it will stop or fail to find them.
Is create-stories better than a simple task list?
It depends on the job. The skill is designed for structured stories from a design epic with traceable criteria. For a quick one-off task list it is heavier than you need, since it asks multiple questions and writes several files.
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-stories if you work in the Claude Code Game Studios structure and want consistent, traceable acceptance criteria across an epic's backlog. Pick something lighter if you only need a quick task list, or if your project does not keep epics, design documents and architecture decision records in the expected folders.
Download the skill