Ralph Wiggum Loop

Name: Ralph Wiggum Loop
Author: skrptiq

Developer Official v1.0.23

Autonomous development loop — pick a task, implement, test, commit, repeat until done

by skrptiq

autonomous claude-code development loops testing

workflow Updated 16 July 2026 12 nodes 2 downloads

What's included 12 nodes

1 workflow 4 skills 4 prompts 3 assets

workflow (1)

Ralph Wiggum Loop

Autonomous development loop — pick a task, implement, test, commit, repeat until all tasks pass

skills (4)

Completion Report

Produces a summary of what the autonomous loop accomplished across all iterations

Implementation

Implements the selected task — writes code, runs tests, and commits the changes

Task Selection

Reads the progress tracker and selects the next incomplete task based on dependencies and priority

Test Verification

Verifies implementation correctness and determines whether the loop should continue, pass, or fail

prompts (4)

Implement Task

Implements the selected task — writes code, runs tests, commits

Report Completion

Produces a summary report of what the autonomous loop accomplished

Select Task

Reads the progress tracker and selects the next task to implement

Verify Tests

Evaluates implementation correctness and decides whether the loop should continue, pass, or fail

assets (3)

Claude Code Configuration

Recommended CLAUDE.md conventions and settings for autonomous loop operation

Loop Runner Script

Reference bash script for running the Ralph Wiggum loop with Claude Code CLI

Progress Tracker Template

Template for the progress file that tracks task completion across loop iterations

Requirements 1 service, 3 permissions, 1 data type

Services

LLM Service

Permissions

filesystem:read
filesystem:write
shell:execute

Data Handling

source code

Security Passed

All checks passed v1.0.23 · scanner v3.3.0

Detected

Services: llm-service
Permissions:
Data Handling:

Full report →

Version history 15 releases

v1.0.23 latest 16 July 2026

GH#845 — republish with American English (en-US) content, completing the source-only GH#805 flip that never reached the Hub. Copy only — no functional or behaviour change.

v1.0.22 2 July 2026

GH#745 — declare per-step `output: {name, type}` on every execution step (selected_task/text, implementation/text, test_results/decision, completion_report/text). Lights up the #744 rich flow-map with named, typed outputs. Content-only; no bindings or logic changes.

v1.0.21 7 June 2026

Fix-forward after Row 3b v1.0.20 publish failure. The v1.0.20 per-skrpt CI's "Register version with Hub API" step failed because the consumer's source `manifest.id` (a7f2d91c…) did not match the D1 catalogue row's id (65e9ba39…) — a legacy drift from before Action 6 (`0bcc5ae0`) made publish-skrpt.mjs Step 2 INSERT use `manifest.id` for the D1 id column. v1.0.21 reconciles the source `manifest.id` to the catalogue authoritative value (Row-5-equivalent for consumers) and republishes. Per Adj-1: no re-tag of v1.0.20; the orphaned GitHub release artefact stays inert (no D1 versions row, no consumer pinned it).

v1.0.19 22 May 2026

Removed spurious glob-pattern directory that triggered scanner warning.

v1.0.15 19 May 2026

Signature fix — RELEASE_NOTES.md now included in integrity checksum.

v1.0.14 19 May 2026

Initial catalogue release with full structural and content-quality validation. All scanner checks pass.

v1.0.13 18 May 2026