agentleFS
whinstudiowhinstudio/skills/game-balance/SKILL.md

game-balance

Skill0 starsChanged 3 days agoMITView on GitHub

Use when a Unity 6 mobile game needs its economy or difficulty written down, or someone asks whether the economy or difficulty is balanced, whether prices, rewards, wheel odds or ad limits are fair, or whether the first minutes are too hard. Also for a balance check of GameEconomy.json against stated targets.

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the data and the targets, then follow this order.

    line 31

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

game-balance

Three modes. economy writes design/economy.md, difficulty writes design/difficulty.md, check compares the real data with the targets in those files and reports. Only check save (or "write it down") writes qa/balance-<date>.md. Nothing else is written.

Never invent a target. A target comes from the user, from design/brief.md (Session length, Revenue) or from a filled template. If there is none, say so and ask, or leave the cell as (unset). A (proposal) value is not a target until the user accepts it; when nobody can answer (an unattended run), it does not count as a target.

Where the data is

  1. Assets/_Project/Compass/Resources/GameEconomy.json (optional): currency, boosters, extras (supply, wheel, login), shop. Field meanings: economy-model.
  2. Tuning assets of the game (ScriptableObjects or config classes: speeds, spawn rates, lives, level data), if the project has them.
  3. design/brief.md, design/economy.md, design/difficulty.md.

No GameEconomy.json and no stated targets: give the verdict NOT ASSESSED — NO DATA and one line offering the template, even when the game has tuning assets. Do not estimate.

Mode: economy

Fill the template in economy-model into design/economy.md. Values that exist in GameEconomy.json are copied, not re-invented; targets are the user's words or (unset). If design/economy.md exists, show the changes and ask before overwriting.

Mode: difficulty

Fill the template in difficulty-curve into design/difficulty.md. Session length comes from design/brief.md. Levels, speeds and lives come from the tuning assets; a value you cannot find stays (unset).

Mode: check

Read the data and the targets, then follow this order.

1. Facts that need no target. Report only what the file proves:

  • a booster has a price but the file lists no way to get currency: no supply with freeCoins above 0 or ad rows, no wheel slice or login day with coins, no shop row with coins. A game's own code (level rewards via economy.Earn) may earn it: say the file shows no source and ask, do not call the price unreachable;
  • a currency source (supply, wheel coins, login coins) in a file where no booster has a price and there is no custom extra: nothing to spend on;
  • a shop row with no coins and no boosters that is not remove_ads;
  • a booster with start 0 and no way to get it (no price reachable, not in a shop row, not a wheel or login reward).

2. Compare with targets. For every value in the template that has a stated target, write the value, the target, and within or outside, with the gap. Gap: for a range, measure against the nearer bound (bounds count as within); for "at most" or "at least", against that limit; for a single value, against it. Gap percent = distance divided by the target. A target of 0 has no gap percent: any other value is outside it and counts as UNBALANCED. A value without a target is listed as "no stated range, not judged". A target the file cannot answer (for example minutes to the first purchase, when earnings live in game code) is listed as "not computable from the file". Neither counts towards the verdict. Do not guess what is typical. The file holds supply.freeHours (an interval) and wheel.freeEvery (runs or won levels), so for "free spins or free supplies per day" the supply part is 24 / freeHours and the wheel part is not computable.

3. Fairness rules, checked against the data:

  • No pay-to-win, as far as the file shows: a booster, or the currency itself, reachable only through shop rows (no start, no wheel or login slice, no price in currency) is a finding. What the booster does in play is game code and is not judged here.
  • Rewarded ads give currency only through an extra (supply.adCoins, wheel.adSpinsPerDay, a custom extra). The daily cap is in the template's code, not the file. Ad exposure per day counts adCoins rows plus adSpinsPerDay only; in-run booster ads are per run and are not counted (say so). Compare with a stated limit, if any.
  • Wheel odds are visible: the template's "i" button lists every slice's odds. A game that replaced the wheel with its own code: check its odds are shown. Wheel and login booster slices are listed but not valued in coins.
  • No timer pressure on essential items: supply.freeHours and countdowns apply to extras, never to something the player needs to continue. A "limited time" label on a shop row (tag is a discount percent) must not be the only way to get an item the game needs.

Difficulty is a separate check against design/difficulty.md: difficulty-curve has its rules; its verdict uses the same list below, on its own values and is reported on its own Difficulty verdict line.

4. Verdict, exactly one; the first that matches:

  • NOT ASSESSED — NO DATA: no file, or no value has a target, or the file does not parse. Name the reason. Facts from the file are still listed. With no file and no targets give only the verdict and one line offering the template, no full report.
  • UNBALANCED: any value outside its target by more than 50% of the target.
  • CONCERNS: any value outside its target by 50% or less, any step 1 fact, or any fairness finding.
  • BALANCED: at least one value judged, all within target, no fact, no fairness finding.

Two agents with the same data must reach the same one.

Output

# Balance check <date>
**Verdict:** <one of the four, for the economy>
**Difficulty verdict:** <one of the four; only when difficulty was checked>
## Facts from the data
## Against targets
| Value | Now | Target | Result |
## Not judged
- <value>: no stated range, not judged
- <value>: not computable from the file
## Fairness
## What to change first (at most 3, each tied to a value)

Report findings in the conversation; write the file only for check save.

Boundaries

Mobile session numbers only: minutes per session, runs to the first purchase, ad rows per day. No live-service measures. Do not edit GameEconomy.json or any game file; propose the change and let the user make it (or have the file regenerated by whatever wrote it). A GameEconomy.json that is written must come with strings in every language for each booster (name and description), the currency name and each custom extra, or EconomyStringsTests fails.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.