chief-wiggum

Chief Wiggum Plugin

A fork of the ralph-wiggum plugin for Claude Code, implementing self-referential AI development loops.

Why This Fork?

This fork exists for two reasons:

1. Bug Fixes

Includes fixes from PR #12642:

• Slash commands failed due to multi-line bash being blocked by security checks
• Permission check bug with ```! auto-execute syntax

Additional fixes in v1.1.0:

• Flag-like values (e.g. --completion-promise --iterations) now rejected with helpful error messages instead of silently misinterpreting arguments
• Iteration context now injected into the prompt Claude sees (workaround for Claude Code's systemMessage not being visible to the AI)

Also adds clearer documentation that completion promises must be wrapped in <promise> XML tags.

2. Natural Language Invocation

You can start a Wiggum loop just by asking Claude naturally:

"Wiggum this until the tests pass" "Keep trying to fix the build" "Run in a loop until it works"

The plugin includes a skill file that recognizes these phrases and automatically starts an appropriate loop with sensible defaults.

Installation

/plugin marketplace add jes5199/chief-wiggum
/plugin install chief-wiggum

Then restart Claude Code to load the plugin.

Note: The Claude Code plugin interface can be buggy. If the above commands don't work, ask Claude to directly edit your config files to install or uninstall the plugin.

What is Ralph Wiggum?

The Ralph Wiggum technique is an iterative development methodology based on continuous AI loops, pioneered by Geoffrey Huntley.

Core concept: Feed Claude the same prompt repeatedly. Each iteration, Claude sees its previous work in files and git history, allowing it to iteratively improve until the task is complete.

This plugin implements Ralph using a Stop hook that intercepts Claude's exit attempts and feeds the same prompt back.

Quick Start

/wiggum-loop "Build a REST API for todos. Requirements: CRUD operations, input validation, tests. Output <promise>COMPLETE</promise> when done." --completion-promise "COMPLETE" --max-iterations 15

Or just ask naturally:

"Wiggum it - implement user authentication with tests"

Commands

/wiggum-loop

Start a loop in your current session.

/wiggum-loop "<prompt>" --max-iterations <n> --completion-promise "<text>"

Options:

• --completion-promise <text> - Phrase that signals completion (HIGHLY ENCOURAGED)
• --max-iterations <n> - Stop after N iterations (HIGHLY ENCOURAGED, recommend 10-20)

IMPORTANT: Always specify BOTH options to prevent infinite loops.

/cancel-wiggum

Cancel the active loop.

/cancel-wiggum

Completion Promises

To signal completion, output a <promise> tag with the exact text specified:

<promise>COMPLETE</promise>

The stop hook detects this tag and ends the loop. Always use both a completion promise AND --max-iterations (recommend 10-20) to prevent infinite loops.

Best Practices

Clear Completion Criteria

Build a REST API for todos.

When complete:
- All CRUD endpoints working
- Input validation in place
- Tests passing (coverage > 80%)
- Output: <promise>COMPLETE</promise>

Always Set Both Safeguards

/wiggum-loop "Try to implement feature X" --completion-promise "FEATURE COMPLETE" --max-iterations 15

Always specify both --completion-promise and --max-iterations (recommend 10-20). This prevents infinite loops on impossible tasks.

When to Use

Good for:

• Well-defined tasks with clear success criteria
• Tasks requiring iteration (getting tests to pass)
• Greenfield projects
• Tasks with automatic verification (tests, linters)

Not good for:

• Tasks requiring human judgment
• One-shot operations
• Unclear success criteria

Learn More

• Original plugin: ralph-wiggum (Anthropic's official version)
• Original technique: https://ghuntley.com/ralph/
• Ralph Orchestrator: https://github.com/mikeyobrien/ralph-orchestrator