Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/plausibleventures/lattice/agents-mdgit clone --depth 1 https://github.com/plausibleventures/latticeWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.03095 | $0.03095 |
| Opus 5 | $0.01548 | $0.01548 |
| Sonnet 5 | $0.00619 | $0.00619 |
| Haiku 4.5 | $0.00310 | $0.00310 |
Grade A, and why
lattice AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 220 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Lattice — how to work in this repo
Lattice is a kit for building isometric, deterministic, zero-asset games in TypeScript. Nine small libraries that compose, and a real game built from nothing but them.
This file is the constitution. It is written for agents first and humans second, because agents outnumber humans here. Read it before you touch anything. It is short on purpose.
- What each package is →
.lattice/kit.json(machine-readable),README.md§ The nine packages, and each package's ownREADME.md. The cross-package contracts are indocs/SEAMS.md - What to work on next →
.lattice/tasks.json - Where the build is →
.lattice/state.json - How one work cycle runs →
docs/LOOP.md
The eleven non-negotiables
These are not preferences. A change that breaks one of these is reverted, not debated.
-
Determinism is a feature, not an accident.
Math.random(),Date.now(), andperformance.now()are banned inside every package'ssrc/. Randomness comes from a seededRngpassed in by the caller; time arrives as a parameter. This is enforced bynpm run lint. A game built on Lattice must be able to replay a session from a seed and an input log and land on the same pixel.And it has two tiers, because the language only promises so much. ECMA-262 specifies
+ - * /,Math.sqrt,Math.imuland the bitwise operators exactly. It explicitly does not requiresin,cos,pow,exporlogto be correctly rounded, so two conforming engines may disagree in the last bit.arithmetic promise may reach Tier A + - * /,sqrt,imul, bitwisebit-identical on every engine hashes, save files, replays, anything Tier B sin,cos,pow,exp,log, …correct to within an ulp or so pixels only — never hashed, never persisted Tier B is not banned; a cost curve is
b · r^kand there is no honest way around that. It is required to declare itself: mark the site@tier-band the linter is satisfied. That makes every one of them greppable, so an auditor can ask of each in turn whether it ever reaches a save file.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 220 lines · 3,095 tokens per session scan A e382ba853b2b
lattice AGENTS.md is an instructions file published in the GitHub repository plausibleventures/lattice (35 stars, last pushed 10d ago), licensed MIT. It adds 3,095 tokens to every session, about $0.0155 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
turn-based-game-mcp typescript.instructions.md
TypeScript development patterns and best practices for the turn-based games platform.
roblox-brain AGENTS.md
AGENTS.md instructions for TabooHarmony/roblox-brain, covering agents.md: guide for ai agents and contributors, what this repo is, architecture: progressive disclosure, skill file structure and skill title.
godot-ai AGENTS.md
AGENTS.md instructions for hi-godot/godot-ai, covering agents.md - godot ai, what this project is, architecture, project structure and key conventions.
tevm CLAUDE.md
Claude Code instructions for evmts/tevm, covering tevm monorepo commands and style guide, project overview, forking implementation, json-rpc support and commands.
GameDesignOS AGENTS.md
Instructions for DY-2026/GameDesignOS, covering gamedesignos agent 入口, 默认接法, agent 执行规则, 什么时候用整个项目 and 每个 skill 单独怎么用.
moss AGENTS.md
Instructions for D-Robotics/moss, covering agents.md, 仓库身份, 文档所有权与阅读顺序, 代码规范执行摘要(必须遵守) and 第一次进入仓库.