Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Ralph Wiggum iterative development loop methodology for persistent AI agent work. Implements continuous iteration loops where AI works on tasks until completion, using Archon for state management and context handoffs. Integrates with The Long Run Harness, Spec Kit, and PRP frameworks. Triggers: ralph, wiggum, iteration loop, persistent agent, continuous development.
.claude/skills/majiayu000-ralph-wiggum/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-11 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -69% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -74% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -80% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 15% | 0% |
<domain_overview>
> Philosophy: "I'm helping!" — Rational: Fix the root, not the symptom. ROOT CAUSE SURGERY MANDATE (CRITICAL): Ralph is not a feature developer. He is a surgical specialist for existing logic failures. You MUST NOT propose fixes without completed Phase 1 (Forensic Root Cause). Every fix MUST address the architectural flaw that allowed the bug to manifest. Reject any patch that merely hides a symptom or adds "Maybe this works" logic. </domain_overview> <autonomous_debugging>
Ralph uses the ralph-harness.js to ruthlessly pursue and eliminate error signals.
@debug-mastery to find the bad value origin..maestro/brain.jsonl for historical context on why this logic was built.Run fix attempts through the persistent orchestrator:
bashnode scripts/js/ralph-harness.js "npm test" --elite
</autonomous_debugging> <code_improvement_loop>
Ralph ensures all existing code meets the @clean-code standard.
Before finalizing any code optimization:
bashnode scripts/js/reflection-loop.js
</code_improvement_loop> <recovery_and_pivots>
When basic fixes fail, Ralph triggers intelligent pivots.
</recovery_and_pivots> <audit_and_reference>
@debug-mastery (Investigation) and @clean-code (Standard).</audit_and_reference>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 88,394 | 9,939 | -89% | 1 | 1 | 0% | 8,265 | 1,072 | -87% | 0 | 0 | — |
case-02 | fail→fail | 10,455 | 10,685 | +2% | 1 | 1 | 0% | 358 | 1,096 | +206% | 0 | 0 | — |
case-11 | fail→pass | 60,712 | 1,688 | -97% | 1 | 1 | 0% | 675 | 1,009 | +49% | 0 | 0 | — |
case-03 | fail→fail | 6,139 | 3,019 | -51% | 1 | 1 | 0% | 254 | 1,106 | +335% | 0 | 0 | — |
case-04 | fail→pass | 28,886 | 5,373 | -81% | 1 | 1 | 0% | 5,162 | 1,589 | -69% | 0 | 0 | — |
case-05 | fail→pass | 38,178 | 5,559 | -85% | 1 | 1 | 0% | 6,055 | 1,583 | -74% | 0 | 0 | — |
case-06 | fail→pass | 39,418 | 4,857 | -88% | 1 | 1 | 0% | 7,290 | 1,446 | -80% | 0 | 0 | — |
case-07 | fail→pass | 15,199 | 8,102 | -47% | 1 | 1 | 0% | 1,635 | 1,877 | +15% | 0 | 0 | — |
case-08 | fail→pass | 11,009 | 7,073 | -36% | 1 | 1 | 0% | 1,530 | 1,772 | +16% | 0 | 0 | — |
case-09 | fail→pass | 12,845 | 2,149 | -83% | 1 | 1 | 0% | 1,983 | 1,046 | -47% | 0 | 0 | — |
case-10 | fail→pass | 13,556 | 6,847 | -49% | 1 | 1 | 0% | 1,401 | 1,014 | -28% | 0 | 0 | — |
case-12 | fail→pass | 28,712 | 6,808 | -76% | 1 | 1 | 0% | 2,699 | 1,090 | -60% | 0 | 0 | — |
case-13 | pass→pass | 15,065 | 3,760 | -75% | 1 | 1 | 0% | 2,220 | 1,335 | -40% | 0 | 0 | — |
case-14 | fail→pass | 12,619 | 5,357 | -58% | 1 | 1 | 0% | 1,924 | 1,584 | -18% | 0 | 0 | — |
case-15 | pass→pass | 14,798 | 3,195 | -78% | 1 | 1 | 0% | 1,541 | 1,210 | -21% | 0 | 0 | — |
case-16 | pass→pass | 8,675 | 8,645 | -0% | 1 | 1 | 0% | 458 | 1,247 | +172% | 0 | 0 | — |
case-17 | pass→pass | 9,634 | 7,224 | -25% | 1 | 1 | 0% | 1,405 | 1,068 | -24% | 0 | 0 | — |
case-18 | fail→pass | 8,289 | 2,356 | -72% | 1 | 1 | 0% | 1,405 | 1,115 | -21% | 0 | 0 | — |
case-19 | pass→pass | 13,837 | 7,993 | -42% | 1 | 1 | 0% | 1,369 | 1,140 | -17% | 0 | 0 | — |
case-20 | pass→pass | 9,597 | 10,905 | +14% | 1 | 1 | 0% | 1,465 | 1,707 | +17% | 0 | 0 | — |
case-21 | pass→pass | 9,652 | 6,836 | -29% | 1 | 1 | 0% | 1,573 | 994 | -37% | 0 | 0 | — |
case-22 | fail→pass | 20,324 | 1,929 | -91% | 1 | 1 | 0% | 2,247 | 1,039 | -54% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 19 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +55 percentage points is the difference between those two pass rates over the 19 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.