next-box
Skill59 starsChanged 2 days agoMITView on GitHub
Add the next box to Contents May Vary (games/contents_may_vary.html), the incremental anthology - compare ideas against the ledger, build a pure core with bots, a view in its own art direction, filmstrip critique rounds, update the rulebook, PR.
In this repository5 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- improve-gameSkill
- improve-toolSkill
- new-gameSkill
- new-toolSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 7
Add one new box to Contents May Vary: a complete incremental game, 30 to 90 minutes long, with a mechanic and an art direction unlike every …
- Very specific: line 17
READ CONTENTS_MAY_VARY.md in full: the rules, the contract, the ledger, the ideas and the lessons.
- Distinctive: line 47
WRITE IT DOWN. Fill in the box's row in the ledger and its pacing, move the runners-up into the ideas, add what you learnt to the lessons.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Add one new box to Contents May Vary: a complete incremental game, 30 to 90 minutes long, with a mechanic and an art direction unlike every box before it. Make every decision yourself and explain the important ones in the PR. Do not stop to ask me.
Idea requested (may be empty): $ARGUMENTS
THE BAR Each box has to stand on its own as a game someone was paid to make, and it has to surprise a player who has opened every box before it.
- READ
CONTENTS_MAY_VARY.mdin full: the rules, the contract, the ledger, the ideas and the lessons. Runnpm run test:cmvand keep it green from here on. Claim the page withpython util/claim.py games/contents_may_vary.html, so no other session works on it at the same time; stop if the claim fails. - CHOOSE. First read What the owner plays in CONTENTS_MAY_VARY.md and answer its checklist for every candidate; a candidate with a dull answer is out. Even with an idea requested, write down at least eight pairs of mechanic and art direction and score each for surprise, depth, fit to 30 to 90 minutes, and distance from the ledger's seven axes (rule 3). Pick one; name it (id, title, handling note). Two lines in the PR on why this box, and why after the last one.
- THE RULES FIRST. Write the core in your scratchpad:
init,step,act,done,progress,milestones, an eager and a casual bot with a person's limits. Tune withKIT.simulateacross several seeds until both bots land in the window with no cliff and no dead stretch. - THE LOOK. Derive the art direction (medium, palette, type, motion, sound, the interface metaphor) from the box's theme, mock up one frame at phone and desktop size, and have a critic subagent that has not seen your reasoning score it (first impression, clarity, art direction, phone). Revise until every score is at least 8. Decide how the box teaches itself in its first thirty seconds and how it ends with a box of its own (rule 8).
- BUILD the view as a
<template data-world>and splice it, with the core, above the marker ingames/contents_may_vary.html. Add the id toSHIPPEDintest/contents_may_vary.mjs, and a suite for the box if one of its rules could break quietly. Keep within the budget (rule 2). 5b. CLOSE UP. List every prop inK.debug.propsand runnode util/cmv_props.mjs <id>; look at every tile at full size against Drawing without glitches in CONTENTS_MAY_VARY.md (one outline per thing, back to front, everything held, light that fades) and give the tiles to a critic with that checklist. Record each transition (the ending, the box opening, a level change) as video and check it frame by frame at 80 ms. Give the box arevisit(S)if carrying on after its end means anything. - PLAY AND CRITIQUE. With
?world=<id>and&at=<minutes>, drive the box with real input in headless Chromium on a phone and a desktop. Record filmstrips (8 to 12 frames) of the first minute, a typical action, each big moment and the ending, and give them to a fresh critic, now also scoring feel. Fix the biggest issues and repeat until every score is at least 8, for at most four rounds; if a score stays lower, say so in the PR. - WRITE IT DOWN. Fill in the box's row in the ledger and its pacing, move the runners-up into the
ideas, add what you learnt to the lessons. Set the page's
updateddate indata/games.json(CLAUDE.md, "Changing an existing game or tool"), commit, push, open a PR, and give me the filmstrips and the critic's last scores.
One box per session.
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
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.

