Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Multi-agent autonomous startup system for Claude Code. Triggers on "Loki Mode". Orchestrates 100+ specialized agents across engineering, QA, DevOps, security, data/ML, business operations, marketing, HR, and customer success. Takes PRD to fully deployed, revenue-generating product with zero human intervention. Features Task tool for subagent dispatch, parallel code review with 3 specialized reviewers, severity-based issue triage, distributed task queue with dead letter handling, automatic deploy
.claude/skills/dokhacgiakhoa-loki-mode/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 516% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 45% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 50% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 20% | 0% |
> Version 2.35.0 | PRD to Production | Zero Human Intervention > Research-enhanced: OpenAI SDK, DeepMind, Anthropic, AWS Bedrock, Agent SDK, HN Production (2025)
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 10,003 | 56,062 | +460% | 1 | 1 | 0% | 1,493 | 9,192 | +516% | 0 | 0 | — |
case-02 | fail→pass | 22,220 | 28,366 | +28% | 1 | 1 | 0% | 4,336 | 6,294 | +45% | 0 | 0 | — |
case-03 | fail→fail | 21,395 | 8,270 | -61% | 1 | 1 | 0% | 4,063 | 1,419 | -65% | 0 | 0 | — |
case-04 | pass→pass | 11,751 | 6,052 | -48% | 1 | 1 | 0% | 1,614 | 1,963 | +22% | 0 | 0 | — |
case-05 | pass→pass | 8,927 | 7,391 | -17% | 1 | 1 | 0% | 1,697 | 2,034 | +20% | 0 | 0 | — |
case-06 | pass→pass | 8,187 | 5,528 | -32% | 1 | 1 | 0% | 1,296 | 1,820 | +40% | 0 | 0 | — |
case-07 | pass→pass | 10,686 | 12,632 | +18% | 1 | 1 | 0% | 1,815 | 3,165 | +74% | 0 | 0 | — |
case-08 | fail→pass | 16,281 | 18,643 | +15% | 1 | 1 | 0% | 2,859 | 4,278 | +50% | 0 | 0 | — |
case-09 | fail→pass | 15,103 | 9,350 | -38% | 1 | 1 | 0% | 2,095 | 2,314 | +10% | 0 | 0 | — |
case-10 | pass→pass | 9,209 | 12,156 | +32% | 1 | 1 | 0% | 1,452 | 2,698 | +86% | 0 | 0 | — |
case-11 | pass→pass | 18,818 | 22,155 | +18% | 1 | 1 | 0% | 2,953 | 4,795 | +62% | 0 | 0 | — |
case-12 | pass→pass | 8,879 | 10,083 | +14% | 1 | 1 | 0% | 1,279 | 2,348 | +84% | 0 | 0 | — |
case-13 | fail→pass | 12,602 | 7,065 | -44% | 1 | 1 | 0% | 1,762 | 2,107 | +20% | 0 | 0 | — |
case-14 | pass→pass | 15,977 | 17,898 | +12% | 1 | 1 | 0% | 2,530 | 4,204 | +66% | 0 | 0 | — |
case-15 | pass→pass | 12,168 | 11,670 | -4% | 1 | 1 | 0% | 1,956 | 2,647 | +35% | 0 | 0 | — |
case-16 | pass→pass | 10,424 | 4,813 | -54% | 1 | 1 | 0% | 1,627 | 1,579 | -3% | 0 | 0 | — |
case-17 | pass→pass | 18,022 | 14,452 | -20% | 1 | 1 | 0% | 2,719 | 3,122 | +15% | 0 | 0 | — |
case-18 | pass→pass | 20,406 | 31,988 | +57% | 1 | 1 | 0% | 3,595 | 6,829 | +90% | 0 | 0 | — |
case-19 | pass→pass | 17,320 | 20,183 | +17% | 1 | 1 | 0% | 2,637 | 3,998 | +52% | 0 | 0 | — |
case-20 | pass→pass | 15,648 | 20,690 | +32% | 1 | 1 | 0% | 2,406 | 3,663 | +52% | 0 | 0 | — |
case-21 | pass→pass | 15,621 | 10,299 | -34% | 1 | 1 | 0% | 2,406 | 3,013 | +25% | 0 | 0 | — |
case-22 | pass→pass | 17,533 | 14,933 | -15% | 1 | 1 | 0% | 3,472 | 3,786 | +9% | 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 21 counted toward the lift figure. The other 1 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 +23 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.