Guide · Agent skill
How to use the Review All Game Design Docs agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The review-all-gdds skill is an agent skill for Claude Code that reads every system game design document (GDD) in a game project at the same time. It looks for problems that only show up when the systems are viewed together, such as contradictory rules, stale cross-references, dominant strategies and drift away from the game's pillars. This guide explains what the skill does, how to install it, how to run it with each focus mode, and what it depends on before you rely on it.
Review All Game Design DocsDonchitos/Claude-Code-Game-Studios · MIT Download the skill Quick start
- 1Confirm your project structure. Check that your project uses the .claude folder structure from the Claude-Code-Game-Studios repository, with project.yaml and design/registry/entities.yaml in place.
- 2Add the helper scripts. Make sure .claude/hooks/yaml-helper.sh and .claude/scripts/review-scope.sh exist, since they are not included in the skill folder.
- 3Install the skill folder. Download the ZIP from this page and unzip it into ~/.claude/skills/ or the project's .claude/skills/ folder, or upload it under Settings, then Capabilities, then Skills in the Claude apps.
- 4Pick a focus mode. Choose full, consistency, design-theory or since-last-review depending on how much of the review you need and how much time you have.
- 5Run the review. Invoke the skill by name with your focus mode. Wait for the full analysis to be shown before you respond to any prompts.
- 6Read the verdict and issues. Check the PASS, CONCERNS or FAIL verdict, then go through the blocking issues first and verify each cited section.
- 7Confirm the report and flags. Approve or decline writing the report and flagging affected GDDs. Make the design decisions yourself.
What does the review-all-gdds skill make the agent do?
An agent skill is a folder of instructions that Claude follows when you invoke it. This one tells the agent to load all of a project's system GDDs, then run two reviews that cannot be done one document at a time. The first is a cross-GDD consistency review: contradictions, stale references and ownership conflicts between documents. The second is a game design holism review: dominant strategies, economy imbalance, cognitive overload, pillar drift and competing progression loops.
The skill is distinct from a single-document review. It checks the relationships between GDDs, not whether one GDD is complete on its own. The output is a written review report with a PASS, CONCERNS or FAIL verdict. The skill can also save that report and flag the affected documents, but only after confirming with you.
- Contradictory rules between systems
- Stale cross-references and ownership conflicts
- Dominant strategies and competing progression loops
- Economic imbalance and cognitive overload
- Drift away from the stated game pillars
Install the review-all-gdds skill
Download the ZIP for the skill from this page. You can install it in either of two ways. For a personal setup, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. Claude Code picks up skills from these locations.
If you use the Claude apps instead, upload the skill under Settings, then Capabilities, then Skills. Installing the folder is only part of the setup. The skill also expects the surrounding Claude-Code-Game-Studios folder structure, described in the requirements section below.
- Option A: unzip into ~/.claude/skills/ for your user account
- Option B: unzip into the project's .claude/skills/ folder
- Option C: upload it in the Claude apps under Settings, then Capabilities, then Skills
How do you trigger the skill with a request?
The skill is user-invocable and takes an optional focus argument. Run it with no argument for a full review, which includes both the consistency and design theory passes. Add one of the focus modes to run a narrower check that finishes faster or targets recent changes.
If you prefer plain language, you can ask Claude to review your GDDs for contradictions or design problems. The skill's description covers those topics, so the agent may pick it up that way, but the documented way to start it is by its name with a focus mode.
- full (or no argument): consistency and design theory passes together
- consistency: cross-GDD consistency checks only, faster
- design-theory: game design holism checks only
- since-last-review: only GDDs modified since the last review report, using git
/review-all-gdds consistencyRun a review and read the output
Run the skill before architecture work starts, after all MVP-tier GDDs are approved, or after any GDD is significantly revised mid-production. The skill loads the documents, runs its checks, and then shows the full analysis before it asks for any action. Each issue should cite the exact GDD, section and text involved.
The skill distinguishes blocking issues from advisory ones. It flags contradictions and gives options, but it does not decide which GDD is correct. You make those design decisions. The verdict is still printed in autonomous mode, but the closing handoff prompt is skipped.
- Read the verdict first, then the blocking issues
- Treat advisory items as optional improvements
- Check each cited GDD section against the source yourself before changing anything
- Confirm before the report is written or GDDs are flagged
Requirements and what the skill depends on
The skill needs Claude Code with the .claude folder structure from the Claude-Code-Game-Studios repository. It also references helper scripts that are not included in this skill folder: .claude/hooks/yaml-helper.sh and .claude/scripts/review-scope.sh. Without them, the skill will fail or behave unpredictably.
Other project files it reads include project.yaml, design/registry/entities.yaml, the .claude/docs/ guidance files and docs/engine-reference/. The since-last-review mode needs git. The skill is designed for a multi-agent game studio setup and runs its phases with game-designer sub-agents.
- Claude Code with the Claude-Code-Game-Studios .claude folder structure
- Helper scripts: .claude/hooks/yaml-helper.sh and .claude/scripts/review-scope.sh (not included in this folder)
- Project files: project.yaml, design/registry/entities.yaml, .claude/docs/ and docs/engine-reference/
- git, for the since-last-review mode
Safety notes: scripts and network access
The skill runs scripts. Its permitted tools include bash calls to yaml-helper.sh and review-scope.sh, and it can read, search and write files. Read the scripts before you run the skill on a project you care about, since they are not part of this folder and you should know what they do.
According to the source, the skill does not use the network. Its writes go to the review report and to the flags on affected documents, and it confirms before writing.
- Runs scripts: yes (yaml-helper.sh and review-scope.sh)
- Uses network: no
- Writes the report and flags GDDs only after you confirm
- Review the helper scripts yourself before first use
Tips for getting useful results
Use the narrower focus modes when you only need part of the picture. Use since-last-review after a single system changes mid-production, because it shows what else that change may affect. Keep the registry and project files current, since the skill reads them.
Treat the output as a checklist for a human review, not as a final sign-off. The findings are generated by an AI and can be incomplete or wrong, so check every cited issue against the source document.
- Start with consistency if you are short on time
- Use since-last-review after revising one system
- Keep design/registry/entities.yaml and project.yaml up to date
- Verify each cited issue against the GDD before acting on it
What people use it for
- Checking that approved MVP GDDs agree on shared rules before architecture work begins
- Finding what else a mid-production revision to one system affects, using since-last-review
- Running a quick consistency pass across documents without the deeper design theory checks
- Spotting an economy where one currency or loop outweighs the others, as part of a design theory pass
- Checking whether later systems still serve the game's stated pillars
Frequently asked questions
Is the review-all-gdds skill free to use?
The skill is published under the MIT licence, as listed on this page. The source does not describe pricing, and you still need Claude Code and the Claude-Code-Game-Studios folder structure to run it.
How do I install the review-all-gdds skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for your user account, or into a project's .claude/skills/ folder. You can also upload it in the Claude apps under Settings, then Capabilities, then Skills.
What do I need before the review-all-gdds skill will work?
You need Claude Code with the .claude folder structure from the Claude-Code-Game-Studios repository. The helper scripts .claude/hooks/yaml-helper.sh and .claude/scripts/review-scope.sh are not included in the folder and must be present, along with project.yaml and design/registry/entities.yaml. The since-last-review mode also needs git.
Can I use this skill on a project with only one design document?
No. The skill is built to review the relationships between documents, so it is not suitable for a single GDD. For one document, use a single-document review instead, such as the separate /design-review command the skill description refers to.
Does the review-all-gdds skill use the internet or run scripts?
According to the source, it runs scripts, specifically the yaml-helper.sh and review-scope.sh helpers, and it does not use the network. Read those scripts before running the skill on important work.
How is this different from a normal design review?
A normal design review checks one GDD for internal completeness. This skill checks how GDDs relate to each other, looking for contradictions, dominant strategies, economy imbalance and pillar drift that only appear when systems are viewed together.
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
This skill suits teams using the Claude-Code-Game-Studios setup who have several approved GDDs and want a cross-system check before architecture work, or who need to see what a mid-production change affects. It is not a good fit for single-document projects, for projects missing the surrounding .claude scripts and folders, or as a final design sign-off without human review.
Download the skill