nirvana squads creator lite
Token-optimized squad generator — 5-phase pipeline with live tech verification, context-manifest pattern, shared macros, and consolidated agents. ~70% fewer tokens than v1.
npx squads add Renat0z/uploads/nirvana-squads-creator-lite -yNirvana Squads Creator Lite
The squad generator that uses ~70% fewer tokens than v1.
A 5-phase pipeline with context-manifest pattern, live tech verification, shared macros, and consolidated agents.
/nscl
Ever wasted hours building squads by hand?
You know the drill.
Create an agent. Then another. Write tasks, wire up cross-refs, validate YAML, realize you forgot a required field, redo everything.
3 hours later, you have a squad that might work.
What if I told you that you could turn a single natural-language sentence into a complete squad — validated, optimized, and ready to deploy — in minutes?
Before and After
| 🐌 Manual | ⚡ NSCL | |
|---|---|---|
| Average time | 2–4 hours | ~5 minutes |
| Cross-ref errors | Frequent | Zero (automatic validation) |
| Token usage | High (re-reads) | ~70% lower (context-manifest) |
| Redundant agents | You don't even notice | AgentDropout eliminates them |
| AIOS validation | Manual, partial | 7 categories, automatic |
| Multilingual README | "I'll do it later" | 6 languages, generated in-pipeline |
How It Works
┌─────────────────────────────────────────────────────────┐ │ YOUR OBJECTIVE │ │ (a single natural-language sentence) │ └──────────────────────┬──────────────────────────────────┘ │ ▼ ┌─────────────────────────┐ │ 🔍 PHASE 1 — ANALYZER │ │ Decomposes objective │ │ into structured │ │ architecture (JSON) │ └────────────┬────────────┘ │ context-manifest.json ▼ ┌───────────────────────────────┐ │ 🔬 PHASE 2 — TECH VERIFIER │ │ Verifies APIs, libs & │ │ services against live sources│ └───────────────┬───────────────┘ │ verified manifest ▼ ┌────────────────────────────────┐ │ 🏗️ PHASE 3 — COMPONENT GEN │ │ Generates agents, tasks, │ │ configs & workflows │ │ in a single pass │ └───────────────┬────────────────┘ │ ▼ ┌───────────────────────────────┐ │ ✅ PHASE 4 — QUALITY GATE │ │ AgentDropout + AIOS │ │ Validation (7 categories) │ └───────────────┬───────────────┘ │ ▼ ┌──────────────────────────────┐ │ 🚀 PHASE 5 — DISTRIBUTOR │ │ READMEs (6 langs) + Deploy │ │ + Publish squads.sh │ └──────────────────────────────┘
Why ~70% Fewer Tokens?
Three architectural decisions make all the difference:
- Context-manifest pattern — The Analyzer produces 1 structured JSON. Downstream agents read that single file instead of re-reading multiple files 4x each.
- Shared macros — Common rules live in . Zero duplication across agents.
/shared
- Single-pass generation — The ComponentGenerator produces everything in one pass. No double-reads, no rework.
Agents
| Name | Archetype | Role | |
|---|---|---|---|
| 🔍 | Analyzer | Guardian | Decomposes objectives into structured architecture |
| 🔬 | TechVerifier | Guardian | Verifies APIs, libs & services against live sources |
| 🏗️ | ComponentGenerator | Builder | Generates all squad components in a single pass |
| ✅ | QualityGate | Guardian | Optimizes (AgentDropout) and validates (7 AIOS categories) |
| 🚀 | Distributor | Flow_Master | Multilingual READMEs, deploy, and publishing |
Tasks
| Task | Responsible Agent | Atomic Layer |
|---|---|---|
| Objective Analysis | 🔍 Analyzer | Decomposition → context-manifest.json |
| Tech Stack Verification | 🔬 TechVerifier | APIs, libs, services → live verification |
| Component Generation | 🏗️ ComponentGenerator | Verified manifest → agents/tasks/configs/workflows |
| Validation & Optimization | ✅ QualityGate | AgentDropout → 7-category validation |
| Distribution & Deploy | 🚀 Distributor | READMEs 6 langs → Deploy → Publish |
Workflows
| Name | Pattern | Description |
|---|---|---|
full-pipeline | Sequential | Complete pipeline: analysis → generation → validation → distribution |
validate-only | Sequential | AIOS validation only for an existing squad |
Commands
| Command | What it does |
|---|---|
/nscl | Starts the full pipeline (prompts for objective) |
/nscl:validate | Validates an existing squad (7 AIOS categories) |
/nscl:analyze | Runs the analysis phase only |
/nscl:generate | Generates components from an existing manifest |
/nscl:distribute | Generates READMEs and publishes |
AIOS Validation — 7 Categories
The QualityGate automatically checks:
| # | Category | What it validates |
|---|---|---|
| 1 | Manifest | squad.yaml is complete and well-formed |
| 2 | Directory | Folder structure follows AIOS spec |
| 3 | Agent Format | Required fields, valid archetypes |
| 4 | Task Format | Inputs/outputs, correct agent refs |
| 5 | Cross-Refs | Every cross-file reference resolves |
| 6 | YAML Syntax | Zero parsing errors |
| 7 | Tech Verification | Verified dependencies, no REJECTED items |
Tech Stack
| Technology | Usage |
|---|---|
| AIOS Core ≥ 2.1.0 | Squad execution runtime |
| Claude Code | AI engine for agents |
| YAML | Configuration format |
| JSON | Context-manifest pattern |
| Shields.io | Dynamic badges |
FAQ
Can I use it with any objective?
Yes. The Analyzer accepts any natural-language description. The more specific the objective, the more precise the generated architecture. A sentence like "squad for SaaS customer onboarding automation" is more than enough.
Can AgentDropout remove agents I want to keep?
AgentDropout only merges agents whose responsibilities overlap significantly. It never removes functionality — it only consolidates redundancy. The result is a leaner squad that does exactly the same thing with fewer tokens.
What's the difference from Nirvana Squads Creator (v1)?
~70% fewer tokens. v1 used separate agents that re-read the same files multiple times. Lite uses context-manifest (1 shared JSON), zero-duplication macros, and single-pass generation. Fewer agents, fewer tasks, same result.
Does it work with existing squads?
Yes. Use
/nscl:validate
What languages is the README generated in?
Português (BR), English, Español, العربية (Arabic), हिन्दी (Hindi), and 中文 (Chinese).
Created by Renato Medeiros · MIT License
squad-creator · meta-tool · aios · multi-agent · claude-code · pipeline · automation · token-optimized · lite
Reviews
0 reviewsNo reviews yet. Be the first to review this squad!
More from Renat0z
brainstormind
Workflow Diverge+Converge — swarm de 24 agentes gera 200+ ideias, filtra Top 3, depois refina o melhor insight em design validado. Pipeline de 6 fases com gate interativo, context-manifest pattern, shared macros. ~65% menos tokens que abordagem naive.
swarm-tree-orchestration
Recursive tree decomposition orchestrator squad — decomposes complex tasks into a tree where each leaf has max 2 sub-tasks, plans the complete tree before execution, launches waves via TaskCreate, and integrates results. Inherits Context Factory, UxIxR scoring, circuit breaker, and fitToBudget.
win proposal deal
Propostas comerciais que fecham — 4 agentes IA analisam seu prospect, desenham 3 opcoes de escopo, precificam com win-rate preditivo e entregam proposta persuasiva pronta em minutos
token-optimizer
Analisa squads AIOS existentes e produz otimizacoes priorizadas por ROI — qualidade, velocidade e economia de tokens — usando TOKEN-OPTIMIZATION-GUIDE.md como base de conhecimento.
design 3d experience
Squad para planejar e construir experiencias 3D web imersivas — do briefing estrategico ao deploy otimizado com Three.js, React Three Fiber, GSAP e scroll-driven narratives
video factory remotion
End-to-end programmatic video creation squad — from briefing to rendered MP4. 5 specialized agents orchestrate storyboard, assets, voiceover+SFX+BGM, Remotion composition (3-layer audio), and render pipeline with post-processing.