ralph-wiggum — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited ralph-wiggum (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Purpose: Autonomous coding loop technique for complex build tasks.
What it does: Runs an AI coding agent (Claude Code, Codex, etc.) in iterative loops — each iteration picks ONE spec/task, implements it, verifies, commits, and loops until done.
Based on: Geoffrey Huntley's Ralph Wiggum technique — fresh context each loop, persistent state via git/files.
Already installed at: ~/.openclaw/skills/ralph-wiggum/
Structure:
ralph-wiggum/
├── SKILL.md
├── scripts/
│ ├── ralph-loop.sh # Claude Code loop
│ └── ralph-loop-codex.sh # OpenAI Codex loop
└── specs/ # Spec files (you create these)Ralph Wiggum commit: d205125cc33745116cce22d883417461174dcde5
Write a spec file in specs/ with:
# Claude Code (recommended)
~/.openclaw/skills/ralph-wiggum/scripts/ralph-loop.sh
# Or OpenAI Codex
~/.openclaw/skills/ralph-wiggum/scripts/ralph-loop-codex.sh
# Limit iterations
~/.openclaw/skills/ralph-wiggum/scripts/ralph-loop.sh 20Each iteration:
1. Ralph reads specs/ and picks highest priority incomplete spec
2. AI implements it completely
3. AI verifies acceptance criteria + runs tests
4. AI outputs "DONE" only if criteria pass
5. Bash loop checks for DONE → next iteration
6. Context cleared, fresh start<promise>DONE</promise> → spec complete, next iteration<promise>ALL_DONE</promise> → all specs complete, loop exitsSpecs: Markdown files in specs/ with acceptance criteria. Lower number = higher priority.
Context Reset: Each iteration starts fresh — no accumulated confusion.
State Persistence: Progress stored in git commits + files.
Completion Verification: AI only outputs DONE when ALL criteria verified + tests pass.
Ralph Wiggum is useful for:
To use with OpenClaw: spawn a subagent with sessions_spawn and run ralph-loop.sh inside it.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.