Godot Idle Clicker: the eval log

We gave Godot Idle Clicker to a coding agent in a new session with no CodePlate context, along with a made-up user's brief ("Bakery Boss"), then sent the requests below one at a time.

5 of 5 requests passed

Run: 2026-10-13
Agent: Claude Sonnet 5.5
The user: Claude Opus 5.5, from the script
Version tested: v1.0.0
On sale now: v1.1.4

This run tested v1.0.0. The version on sale is v1.1.4, which includes the fixes listed at the end of the record and has not had its own eval run.

The made-up user

Written the way a buyer would put it. {ASSETS}, {ART} and {GODOT} stand for the folder with their files and the engine's path; their asset pack isn't published.

The user is a hobbyist who builds with an agent. They don't read code.

Game: Bakery Boss, a cookie idle game.
Art: only a cookie (384×384) and a bakery background (720×1280), in assets/. Keep the template's icons for now.
Platforms: mobile first, web too.

First message

Hi! I want to turn this idle template into "Bakery Boss": you tap a giant cookie and hire grandmas, ovens and factories to bake for you. My cookie and background art are in {ART}. Godot 4.7.2 is at {GODOT}. Can you take a look and get us started?

Interview answers

Currency: cookies. Machines, in order: Grandma, Oven, Mixer, Bakery, Factory, Cookie Farm, Bake Lab, Time Oven.
Keep the template's icons for now. The art is mine; credit "Bakery Boss team".
A bit quicker than the default.
Mobile first, web too.

The requests

Make it Bakery Boss: my cookie and background, the new machine names, cookies instead of gems everywhere.
Add a 9th machine, the "Cookie Portal", after the Time Oven.
Add an upgrade "Bulk Flour" that makes every machine 10% cheaper.
The first ascend should come after about 20 minutes instead of 40.
Add a daily reward: once a day you can claim a gift worth one hour of your cookie production.

The run record

As we wrote it on 2026-10-13, with local paths and internal references taken out. "I" is the person running the eval.

Agent: Claude Sonnet 5.5 (general-purpose subagent), with no CodePlate context. It was given only the user's messages and paths.
User: played by Claude Opus 5.5, following concept.md and requests.md.
Art pack: only a cookie (384×384) and a bakery background (the eval's art pack). Everything else stays a placeholder.

Results

Step	Result	Notes
Interview	pass	It asked 8 focused questions with defaults, and noticed the striped background might hurt readability.
1. Re-theme	pass	Renamed every machine, upgrade, achievement, tab and prestige (Franchise and golden sprinkles), and used "cookies" everywhere. Kept the ids so saves stay compatible, and said so. Dimmed the background for readability. Reused the cookie for the counter icon, the tap particles and the app icon. It also tuned pacing early (prestige threshold 1M → 300K), from the interview answer "a bit quicker".
2. 9th machine	pass	Used the recipe's ×12 cost / ×6 rate rule, and generated a labelled placeholder icon. Moved the all-machines target from 120 to 165 minutes, and explained the alternative (a cheaper portal).
3. Bulk Flour	pass	Added a new `cost_discount` effect to cost, bulk buying and MAX, with validation (0–1), a check and docs. Flagged that the pacing bot undervalues non-rate upgrades.
4. ~20 min Franchise	pass	Moved it from 28m31 to 20m23. Moved the target to 22 minutes. Reported the knock-on effect (more sprinkles per run) and the levers to rein it in.
5. Daily reward	pass	The rule is in `Economy`, with the day passed in, so there's no clock in core. It's saved and loads older saves, the button hides when there's no production, and there are 2 new checks. It explained the UTC-day choice and offered a 24-hour alternative.

The checks went from 25 to 28, all passing. Each request took one agent turn of 2.5–6 minutes.

Template fixes from this run

The screenshot tool can scroll the shop (scroll=end). The agent couldn't see the machine it had just added at the end of the list.
Known limitation, already in the recipe: the pacing bot values upgrades by the rate they add, so discounts and other non-rate upgrades are bought late. That's fine for pacing checks; a smarter valuation can come later.

Takeaway

The idle template's data-driven design (JSON content, a pure Economy, a pacing bot with targets) made every request a clean, local change. It was the smoothest eval so far.

What each request was scored against

Every request: checks PASS, no check weakened silently, docs/project.md updated.

Re-theme
- The cookie replaces the crystal, and the background is replaced.
- Generator names are changed, but their ids are kept (or the agent asked first).
- The user-facing text says "cookies".
- assets.json is updated.
- The agent took a screenshot and looked at it.
9th machine: a new generators.json entry with sensible cost and rate. Pacing was run and the targets still hold, or were adjusted with an explanation.
Bulk Flour: a new effect type (cost_discount) in Content.EFFECTS and Economy.cost() / max_affordable(), with a check and docs.
Faster pacing: target_first_prestige_minutes is about 20. Costs, rates or the prestige threshold were tuned so the pacing check passes, and the before/after was reported.
Daily reward: the rule is in Economy, with the day passed in from Game.clock (no clock in core), saved and loaded. There's UI (a button or tab), checks for once-per-day, the amount and the save round-trip.

← Godot Idle Clicker · Every eval log