Guide · Agent skill
How to use the Scope Check agent skill: detect scope creep
Updated · written with AI from the project's own documentation and checked against our catalogue data
Scope Check is an agent skill from the Donchitos Claude-Code-Game-Studios project that makes Claude act as a scope auditor. You give it a feature, sprint, or milestone name, and it compares the original plan with what has actually been built or committed. This guide explains what the skill does, how to install it, how to trigger it, and where its results are limited.
Scope CheckDonchitos/Claude-Code-Game-Studios · MIT Download the skill Quick start
- 1Download the ZIP. Download the Scope Check ZIP from the page you are reading.
- 2Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use, or into your project's .claude/skills/ folder for one project.
- 3Confirm git and the plan document. Make sure git is available in the project and that a plan document exists in design/ or production/.
- 4Ask for a scope check. Use a natural request or the /scope-check command with a feature name, sprint number, or milestone.
- 5Read the report and verdict. Review the added and removed items, the net change percentage, and the verdict.
- 6Act on the recommendations. Cut, defer, or flag items as suggested, then re-run the check to confirm the verdict improves.
What an agent skill is and what Scope Check makes the agent do
An agent skill is a folder containing a SKILL.md file with instructions that an AI agent can load and follow for a specific task. Scope Check is one such skill. It is read-only, so it reports findings and writes no files.
When you ask for a scope check, the agent finds the original plan document, looks at the code and commit history for the same work, and produces a report. The report lists added and removed items, a percentage change, risk ratings, and suggested cuts or deferrals.
If the plan or the current state cannot be read, the skill returns a NOT ASSESSED verdict instead of a misleading 0%.
- Compares the original plan with what is implemented or in progress
- Lists additions and removals in tables
- Calculates a net scope change as a percentage
- Rates schedule, quality, and integration risk
- Recommends items to cut, defer, keep, or flag for a decision
Install the Scope Check skill
Download the ZIP for this skill from the page you are reading. Then choose one of two installation routes.
For a personal install that works across your projects, unzip the download into ~/.claude/skills/. To limit it to one project, unzip it into that project's .claude/skills/ folder instead.
In the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal: unzip into ~/.claude/skills/
- Single project: unzip into your project's .claude/skills/
- Claude apps: Settings → Capabilities → Skills, then upload
Check what the skill needs before you run it
Scope Check needs two things in your project. First, git must be available, because the skill runs git log to read commit history for the work. Second, you need a written plan document in the design/ or production/ folders that follows the expected layout.
The skill uses no scripts and no network access, according to the skill's listing. It only reads files and runs git log.
- git available in the project
- A plan document in design/ or production/ following the expected layout
- Runs scripts: no
- Uses network: no
Trigger the skill with a natural request or argument
You can ask in plain language, such as asking whether there is any scope creep in a feature. The skill's argument can be a feature name (several words are allowed), a sprint number, or a milestone name.
The source gives these locations for the baseline document. A feature name is matched to design/gdd/[feature].md or a matching file in design/. A sprint such as sprint-3 is matched to production/sprints/sprint-003.md. A milestone is matched to production/milestones/[name].md.
- "Any scope creep on the inventory feature?"
- "Run a scope check on sprint-3"
- "Check scope for the combat milestone"
/scope-check [feature-name or sprint-N]Read the verdict and act on it
The skill assigns one verdict based on net scope change. PASS means on track, with change of 10% or less. CONCERNS means minor creep, with change above 10% up to 25%, manageable with targeted cuts. FAIL means significant creep above 25% up to 50%, or out of control above 50%. NOT ASSESSED means the comparison could not be made.
NOT ASSESSED outranks PASS, because a comparison that never happened has not shown scope is on track. If the baseline is empty or the current state is unreadable, populate the plan or point the skill at where the work lives, then run it again.
For every verdict except NOT ASSESSED, the skill suggests re-running the check after cuts are made to confirm the verdict improves.
- PASS: 10% or less
- CONCERNS: above 10% to 25%
- FAIL: above 25%
- NOT ASSESSED: baseline or current state unreadable
Tips for useful results
The verdict depends on how the plan and the agent interpret the documents, so treat results as advisory. Keep your plan document current and listed as concrete scope items, because a baseline with only headings or placeholders produces NOT ASSESSED.
Use commit messages and git history that refer to the feature, since the skill reads commits related to the work. Items the skill reports should be checked against your own knowledge before you make decisions.
- Keep the plan document specific, with real scope items rather than placeholders
- Make sure the work is in git history for the window being checked
- Treat the percentage as evidence to discuss, not a precise measurement
What people use it for
- Checking whether a sprint has grown beyond its written plan before a milestone review
- Getting a quantified view of a feature's additions and removals for a producer or lead
- Reviewing a small team's design-doc workflow to find work that was never planned
- Deciding which nice-to-have additions to defer to a future sprint
Frequently asked questions
How do I install the Scope Check agent skill?
Download the ZIP from the page, then unzip it into ~/.claude/skills/ for personal use or into your project's .claude/skills/ folder for one project. You can also upload it in the Claude apps under Settings → Capabilities → Skills.
What does the scope-check skill need to work?
It needs git available in the project, because it runs git log to read commit history. It also needs a plan document in the design/ or production/ folders that follows the expected layout. The listing says it runs no scripts and uses no network.
Is the Scope Check skill free?
The source gives the licence as MIT, but it does not state any price. Check the repository for licence terms before you use it.
Why does Scope Check return NOT ASSESSED?
It returns NOT ASSESSED when the baseline plan has no scope items, or when the current state cannot be determined, such as no related files or no commits in the window. This avoids a misleading 0% result. Populate the plan or point the skill at where the work lives, then run it again.
Does Scope Check change my files?
No. The skill is read-only and writes no files. It only reports findings.
How is Scope Check different from just reading my sprint plan?
Scope Check compares the plan against what has been built or committed, using code search and git history. It then produces a quantified report with a percentage change, risk ratings, and suggested cuts or deferrals.
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
Scope Check suits game or software developers running sprints, producers who want an evidence-based view before a milestone, and small teams that keep a written design document. It is a poor fit for projects with no written plan, because it will report NOT ASSESSED, and for anyone who needs a precise objective measurement, since results are advisory.
Download the skill