@kikompg/neuron-cc
v0.2.4
Published
Neuron installer for Claude Code process discovery, capability routing, and department-agent generation.
Maintainers
Readme
Neuron
Process discovery, capability routing, and department-agent generation for Claude Code, coordinated by a central Neuron orchestrator.
Install
npx @kikompg/neuron-cc@latest --claude --globalUpdate
npx @kikompg/neuron-cc@latest --updateMain command
/neuronAdditional direct commands:
/neuron-avaliar-projeto
/neuron-analisar-reuniao
/neuron-gerar-agents-empresa
/neuron-atualizar-agents-empresaWhat it does
- Analyses company context from transcripts or business descriptions
- Identifies relevant departments, challenges, and automation opportunities
- Generates specialised department agents as Markdown files
- Generates a CEO-facing agent that consolidates department improvement needs
- Updates generated agents when new company information appears
- Uses a central orchestrator to control the main execution flow and avoid redundant runs
- Reuses shared skills for process mapping, opportunity analysis, scoping, design, and SOP work
Repository architecture
neuron/
├── agents/
│ ├── core/
│ │ ├── neuron-orchestrator.md
│ │ ├── company-analyzer.md
│ │ ├── opportunity-analyzer.md
│ │ ├── project-evaluator.md
│ │ ├── department-agent-designer.md
│ │ ├── department-agent-updater.md
│ │ └── ceo-agent-designer.md
│ └── generated/
│ └── acme-industrial/
├── skills/
├── company-data/
│ └── acme-industrial/
├── commands/
├── .claude/
├── .claude-plugin/
├── scripts/
└── README.mdCore agents
neuron-orchestrator-> controls the main Neuron flow and decides which agents run nextcompany-analyzer-> identifies company context, departments, and agent candidatesopportunity-analyzer-> identifies operational and automation opportunitiesproject-evaluator-> routes work across available capabilities and fallbacksdepartment-agent-designer-> generates department-specific Markdown agentsdepartment-agent-updater-> updates existing generated agents without duplicationceo-agent-designer-> generates a CEO-facing synthesis agent from all department agents
Reusable skills
- process-mapping
- company-discovery-checklist
- opportunity-identification
- opportunity-scoring
- automation-scoping
- sop-generation
Main workflows
1. Initial meeting diagnosis
Use:
/neuron analisar-reuniao [transcrição ou resumo]Output format:
# Neuron
## Resumo
## Top oportunidades
## Recomendação
## Próximo passoThis step is intentionally short:
- no JSON
- no capability routing
- no full report
1b. Short project evaluation
Use:
/neuron avaliar-projeto [contexto]Output format:
# Neuron Evaluation
## Resumo
## O que já existe
## O que falta
## Fallback agora
## Próximo passo2. Generate company department agents
Use:
/neuron gerar-agents-empresa [descrição da empresa ou transcrição]What it does:
neuron-orchestratorcoordinates the flow- analyses the company
- creates a canonical company memory in
company-data/<company-slug>/company-context.md - identifies relevant departments
- identifies department improvement needs
- creates department agents in
agents/generated/<company-slug>/ - creates
ceo-agent.mdwith the consolidated executive view
3. Update generated agents
Use:
/neuron atualizar-agents-empresa [nova informação da empresa]What it does:
neuron-orchestratorchecks the current pipeline state- reads canonical company memory
- identifies affected agents
- updates company memory first
- updates only the necessary files
- preserves validated information
- avoids duplicate department agents
Generated agent model
Each generated agent should contain:
- YAML metadata block
- name
- canonical key
- department
- purpose
- company context
- when to use
- when not to use
- inputs
- outputs
- responsibilities
- useful tools
- core skills
- optional skills
- skill usage
- department analysis
- update rules
Required YAML metadata:
agent_iddepartmentcompanybase_agentversioncreated_bylast_updated
Each department agent should use this fixed analysis contract:
## Departamento## Problemas encontrados## Melhorias recomendadas## O que pode ser automatizado## O que deve continuar manual## Prioridade## Dependências
Inside ## Prioridade, every agent should score:
impactocomplexidadeurgenciadecisao
These scores should follow:
The CEO agent should then consolidate those scores into:
## Executive Summary## Top Priorities## Quick wins## Cross-Department Issues## Dependencies## Recommended Roadmap## Next Step
Example files:
Company context storage
Reusable company knowledge lives in:
This context is the base for:
- agent generation
- agent updates
- preserving validated information over time
Neuron now treats this file as a canonical company memory, not a loose summary. That memory should separate:
- validated facts
- working hypotheses
- departments
- existing generated agents
- discovery checklist items marked as
confirmed,inferred, orunknown - decisions made
- open questions
Reference template:
Guardrails
- do not create redundant agents
- do not duplicate overlapping departments or roles
- create department agents only when there is clear operational need
- prefer updating an existing agent over creating a new one
- resolve departments to stable canonical keys before creating files
- keep deterministic file names such as
sales-agent.md,finance-agent.md,operations-agent.md
Reference rules:
Example commands
/neuron analisar-reuniao Cliente industrial com vendas, finanças, RH e operações. Sistemas atuais: Gmail, CRM, ERP e folhas de cálculo. [colar transcrição]/neuron gerar-agents-empresa Empresa industrial B2B com equipas de vendas, finanças, RH e operações. Sistemas atuais: Gmail, CRM, ERP e folhas de cálculo. [colar transcrição]/neuron atualizar-agents-empresa Nova informação: operações foi dividida entre logística e planeamento, e finanças passou a incluir controlo de crédito. [colar update]