simpsons-development

Simpsons Development

Five Simpsons-themed self-referential development loops for Claude Code

The Family

Character	Command	Discipline	Phases	Emoji
Lisa Simpson	/lisa-loop	Specification-First Development	6	saxophone
Ralph Wiggum	/ralph-loop	Iterative Development	1	arrows
Bart Simpson	/bart-loop	Adversarial Red Team	6	skateboard
Homer Simpson	/homer-loop	Incident Response & Resilience	6	donut
Marge Simpson	/marge-loop	Quality Governance & Maintenance	6	blue heart

Installation

# From GitHub
claude plugin add https://github.com/reggiechan74/simpsons-development

# Or clone into your plugins directory
git clone https://github.com/reggiechan74/simpsons-development.git

Quick Start

# Lisa: Define specs before building
/lisa-loop Design a rate limiter for the API

# Ralph: Iterate until done
/ralph-loop --completion-promise "DONE" Build the CLI tool

# Bart: Red-team after building
/bart-loop Pen test the authentication API

# Homer: Chaos-test in production
/homer-loop Test the payment service resilience

# Marge: Review before merge
/marge-loop --completion-promise "APPROVED" Review the auth refactor PR

How It Works

Each loop uses the same architecture as the Ralph Loop technique, extended with phase-aware stop hooks:

1. A command (e.g., /bart-loop) loads the character's skill file and calls the shared setup script
2. The setup script creates a state file (.claude/simpsons-loop.local.md) with the character, phase, iteration count, and user prompt
3. Claude works on the task, modifying files and running commands
4. When Claude tries to exit, the stop hook intercepts:
   • Reads the state file to determine current character and phase
   • Checks Claude's last output for a <promise> tag
   • If the promise matches the current phase's promise: advance to next phase
   • If no match: re-feed the same prompt for another iteration
5. After the final phase completes, the loop exits (or continues if --completion-promise is set)

User prompt → Setup script → State file → Claude works → Stop hook
                                                              |
                                          ┌───────────────────┤
                                          │                   │
                                    Promise match?      No match
                                          │                   │
                                   Advance phase      Re-feed prompt
                                          │                   │
                                    Last phase?        Increment iteration
                                          │
                                       Exit loop

Temporal Ordering

When to use which character in your development workflow:

BEFORE --> DURING --> AFTER --> ALWAYS
  Lisa      Ralph      Bart      Marge
                        |
                        v
                      Homer

• Lisa runs BEFORE building — defines specifications, documentation, formal proofs
• Ralph runs DURING building — naive iteration until the task is done
• Bart runs AFTER building — adversarial testing, fuzzing, boundary violation
• Homer runs POST-DEPLOY — chaos engineering, incident response, production hardening
• Marge runs ALWAYS — code review, quality gates, tech debt management

Commands Reference

Loop Commands

All loop commands accept these options:

Option	Description	Default
--max-iterations N	Maximum total iterations across all phases	Unlimited
--completion-promise TEXT	Final exit promise (loop continues after last phase until detected)	Auto-exit
--start-phase N	Start from phase N instead of 1	1

/bart-loop [OPTIONS] PROMPT
/lisa-loop [OPTIONS] PROMPT
/homer-loop [OPTIONS] PROMPT
/marge-loop [OPTIONS] PROMPT
/ralph-loop [OPTIONS] PROMPT   # Note: Ralph has no --start-phase (1 phase only)

Utility Commands

Command	Description
/cancel-simpsons	Cancel the active Simpsons loop
/simpsons-status	Show current character, phase, and iteration
/simpsons-help	Explain the plugin and all commands

Phase Reference

Bart Simpson — Adversarial Red Team