← Back to list

leo-wiggum
by LeomaiaJr
⭐ 0🍴 0📅 Jan 16, 2026
SKILL.md
name: leo-wiggum description: Autonomous AI coding loop with PRD-based task tracking. Use when user says "leo-wiggum", "/leo-wiggum", "start leo loop", "autonomous coding", "run leo", or wants to implement a feature using iterative AI sessions that spawn fresh Claude Code instances. Takes a feature description, breaks it into user stories in a PRD, and runs an external loop where each iteration picks the next incomplete story.
Leo Wiggum - Autonomous AI Coding Loop
External Ralph-style loop that spawns fresh Claude Code sessions to implement features iteratively.
How It Works
- Generate a PRD (
prd.json) with user stories from your feature description - Create
progress.txtfor learnings across iterations - Run external bash loop that spawns fresh Claude Code sessions
- Each session: picks next story → implements → tests → commits → marks done
- Memory persists via git commits,
prd.json, andprogress.txt
Usage
Parse from user input:
- prompt (required): Feature description
- --max-iterations N: Max iterations (default: 10)
- --branch name: Git branch (default:
leo/<feature-slug>)
Step 1: Analyze Codebase
Before generating PRD:
- Read
CLAUDE.mdif exists - Explore relevant code with Glob/Grep
- Understand tech stack and patterns
Step 2: Generate User Stories
Break feature into small stories completable in ONE iteration.
Right-sized:
- Add database column + migration
- Add single UI component
- Update one API endpoint
- Add filter/dropdown
- Write tests for one module
Too big (split):
- "Build entire dashboard"
- "Add authentication"
- "Refactor API"
Step 3: Create prd.json
{
"project": "<project name>",
"branchName": "leo/<feature-slug>",
"description": "<feature description>",
"userStories": [
{
"id": "US-001",
"title": "<title>",
"description": "As a <user>, I want <goal>, so that <benefit>",
"acceptanceCriteria": [
"<criterion 1>",
"<criterion 2>",
"Typecheck passes",
"Tests pass"
],
"priority": 1,
"passes": false,
"notes": ""
}
]
}
Priority 1 = highest. For UI stories, add: "Verify in browser"
Step 4: Create progress.txt
# Leo Wiggum Progress Log
Feature: <name>
Branch: <branch>
Started: <datetime>
## Codebase Patterns
(Patterns added as discovered)
---
Step 5: Show Summary
Display:
- Number of stories
- Story titles with priorities
- Branch name
- Command to run
Step 6: Start Loop
Ask user to confirm, then run:
${CLAUDE_PLUGIN_ROOT}/scripts/leo-wiggum.sh <max_iterations>
CRITICAL: After starting, END response. The script spawns NEW sessions.
Monitoring
- Terminal: iteration progress
progress.txt: learningsgit log: commitsprd.json: story status
Score
Total Score
50/100
Based on repository quality metrics
✓SKILL.md
SKILL.mdファイルが含まれている
+20
○LICENSE
ライセンスが設定されている
0/10
○説明文
100文字以上の説明がある
0/10
○人気
GitHub Stars 100以上
0/15
○最近の活動
3ヶ月以内に更新がある
0/10
○フォーク
10回以上フォークされている
0/5
✓Issue管理
オープンIssueが50未満
+5
✓言語
プログラミング言語が設定されている
+5
○タグ
1つ以上のタグが設定されている
0/5
Reviews
💬
Reviews coming soon