How to use Claude to calibrate a hexcrawl game
Claude Code simulated a hexcrawl's encounter curve and revised color behavior. Human play and player feedback changed the results despite passing tests.
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Inspect the recorded hexcrawl encounter rules and simulate the frequency curve across early and later travel. Show reproducible results against my target interval, then wait for playtest feedback before final tuning. For palette changes, report simulated color-vision results and keep player validation outstanding.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- GlowingOrb
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- Encounter simulation led to changes that were further adjusted by playtesting. Palette and color-filter revisions were still awaiting additional player feedback.
- Human involvement
- GlowingOrb chose rules and features, reviewed increments, played encounter changes and sought feedback from a player with color-vision differences.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- GlowingOrb recorded requirements and bugs before agents planned and implemented increments.
- Claude simulated encounter behavior and measured palette behavior against the reported problems.
- Human play and player feedback drove further corrections despite passing automated tests.
GlowingOrb used Claude Code while developing The Laresian Order, a Godot hexcrawl and tactical RPG. Requirements and bugs lived in an Obsidian vault, and agents worked in separate project copies with access to editor inspection and screenshots through MCP.
One concrete problem was the encounter curve. Reducing the chance of a road fight made early travel too quiet and later travel too busy. An agent simulated game days and showed that the curve's shape, not just its height, needed correction. GlowingOrb supplied a target interval, then played the revised game and adjusted it again when the mathematical result still felt too sparse.
Color accessibility needed a different check. A friend with a red-vision weakness found the unfiltered picture clearer, so the agent measured the palette under simulated color blindness and revised both colors and filter strength. The source explicitly says further player feedback was still needed.
The account describes implementation, simulations and a functioning test-and-review loop, with design and art choices staying human. Automated tests ran before merges, while manual smoke tests still found problems in already-tested code. The result is an iteratively improved game in development, not evidence that a passing suite settles balance or accessibility.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: itch.io devlog · GlowingOrb
Frequently asked questions
What was wrong with encounter frequency?
The curve had the wrong shape, making early and later travel behave differently.
Did simulation settle game balance?
No. GlowingOrb changed the result after playing it.
Was color accessibility confirmed?
No. Palette and filter revisions were still awaiting further feedback.
More details
About this account
GlowingOrb recorded requirements and bugs before agents planned and implemented increments. Claude simulated encounter behavior and measured palette behavior against the reported problems.
Original dated devlog opened and HTML metadata checked. Color accessibility remained under evaluation; AI-assisted drafting of the post is disclosed. No independently audited test or pull-request totals are claimed. Suggested prompt written by Agents In Use; no original task prompt was published in the verified account.
Credit: GlowingOrb — itch.io devlog
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Godot MCP · Obsidian · GitHub
Encounter simulation · Human playtesting
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.