mbutler

Rust Application Development Guide for Coding Agents

Version 1.0 · 8 October 2026

Build ambitious Rust applications through small, verifiable changes. Your responsibility is to leave the application easier to understand and extend after each task, while steadily delivering useful behavior. A successful implementation preserves the user's work, respects the project's boundaries, and gives the next agent enough evidence to continue without reconstructing the entire project.

Read this guide before bootstrapping a project. Then create a short, project-specific AGENTS.md that records the decisions actually made and routes future agents to the relevant sections. Read the complete guide again when substantially changing the architecture; for routine work, use the targeted reading procedure below.

The defaults here suit a solo developer using coding agents with limited context and token budgets. Adapt them to the application's requirements. “Must” identifies a constraint needed to preserve correctness, user d

mbutler

(async () => {
  // 1. Initialize count if not yet defined
  if (typeof globalThis.timeUnitsRemaining !== "number") {
    globalThis.timeUnitsRemaining = 144;
  }

  // 2. Prompt GM for units and activity description
  new Dialog({
    title: "Spend Time Units",
    content: `

mbutler

<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Hinton's Cubes — a working model from The Fourth Dimension (1906)</title>
<style>
:root{
  --paper:#f2ead6;
  --paper-2:#e9dfc4;

mbutler

#!/usr/bin/env python3
"""
Art Object — an Oulipian image machine.

A single-file, stdlib-only generative art tool. The text you give it is not
a caption and not merely a random seed: it is the program. Letter
frequencies, word rhythm, sentence structure, punctuation and entropy are
compiled into a pipeline of image operators. The same text always produces
the same image (its "edition"); a --seed produces numbered variants.

mbutler

Tauri + Rust Senior Engineer Agent Harness
Role & Objective
You are a Senior Staff Rust & Tauri Engineer acting as a continuous pairing partner. Your primary goal is to help build a highly secure, performant, and maintainable desktop application. You prioritize engineering excellence, architectural pragmatism, and strict boundary enforcement over quick, messy fixes.

The user "does not know what they do not know." It is your responsibility to proactively identify anti-patterns, security risks, and technical debt before writing code.

Core Architectural Philosophies

No Premature Abstraction: Do not invent complex abstraction layers for databases, authentication, or state until a pattern naturally repeats. Follow the Rule of Three.

mbutler

AutoResearch is a framework for autonomous, recursive self-improvement. By defining a strict objective metric and a sandboxed testing environment, you can let an AI agent continuously run experiments, keep what works, and discard what doesn't.

When implemented correctly, it effectively replaces hundreds of manual A/B tests with an automated loop that can run 24/7. When implemented poorly, it will confidently optimize your system toward a meaningless metric.

Here is a comprehensive guide to bootstrapping an AutoResearch loop for any measurable task.

---

1. The Core Philosophy

mbutler

import subprocess
import re

# 1. Initialize Memory
messages = [
    {"role": "system", "content": "You are an agent. Write code in ```python filename.py format. I will run it and tell you the result."}
]
max_iterations = 3
current_iteration = 0

mbutler

MISSION: [MISSION NAME]
TYPE: Skill Challenge (Complexity X)
OBJECTIVE: [Short objective]

SUCCESS: [#] / [#]
FAILURES: [#] / [#]

PRIMARY SKILLS
• Arcana DC [ ] → [effect]

mbutler

YESHybrid-v1.1 — Fairy-Stockfish Variant Specification

Overview

YESHybrid-v1.1 is a custom asymmetric Fairy-Stockfish variant played on a 12×8 board. The game emphasizes uneven objectives: White hunts a specific royal target, while Black aims for total elimination or immobilization.

---

Board Configuration

mbutler

Scholarly GIS Research: Specification for Leaflet-Based Mapping Projects

Audience: Development team (experienced web application developers with little or no GIS background).
Purpose: LLM-assisted, spec-driven development of humanities/scholarly mapping projects.
Stack: Leaflet only. Open source tools and standards-based development.

---

1. GIS Concepts for Web Developers