Claude Project
Gojozataro1/Claude_Project
Dashboard de saúde macOS com análise inteligente por Claude e snapshots reversíveis
Ver no GitHubColeção de projetos de engenharia de IA construídos com Claude e a API Anthropic. Inclui um dashboard de operações macOS automatizado que coleta métricas do sistema, envia para análise por Claude e publica recomendações priorizadas em GitHub Pages, com suporte a rollback e múltiplos modos de varredura.
Instalação
npx convez-ai Gojozataro1/Claude_Project1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Monitorar e otimizar saúde do sistema macOS automaticamente
- Auditorias de segurança e vulnerabilidades com IA
- Limpeza inteligente de artefatos de desenvolvimento com rollback seguro
Veredito

Projeto sólido de aprendizado que demonstra integração prática com Claude API, prompt caching e automação real em macOS. Código bem estruturado com agentes modulares, mas ainda em fase de experimento educacional — não está em produção estável.
README
A tradução para o português deste README ainda está na fila.
Claude Project
A collection of AI engineering projects built while learning to work with Claude and the Anthropic API.
Projects in this repo
| Project | Description |
|---|---|
| macOS AI Ops Dashboard | Automated macOS health scanner powered by Claude AI |
| Python Learning Projects | Weekly learning projects: ServiceNow, ticket classification, self-improving agent |
| Otto | Two-page SaaS marketing mockup — pure HTML/CSS/JS |
macOS AI Ops Dashboard
Automated macOS health scanner that collects system metrics, sends them to Claude for analysis, and publishes a prioritised recommendations dashboard to GitHub Pages.
Features
- 4 Scan Modes —
all,dev,vulnerability,ui - Claude AI Analysis —
claude-sonnet-4-6with prompt caching to reduce token costs - Rollback Agent — snapshots system state before any cleanup so you can restore if something breaks
- Post-Cleanup Checker — verifies 6 health dimensions after cleanup (CPU, memory, disk, DNS, services, app integrity)
- GitHub Pages — dashboard auto-published to
docs/index.htmlvia GitHub Actions
Quick start
# Install dependencies
pip3 install -r requirements.txt
# Copy and fill in your API key
cp .env.example .env
# Full scan (all collectors, opens dashboard)
python3 main.py
# Skip Claude AI (no API key needed)
python3 main.py --no-ai
# Security audit mode
python3 main.py --mode vulnerability
# Developer mode — build artifacts, node_modules, IDE processes
python3 main.py --mode dev
# Post-cleanup health check only
python3 main.py --post-check
Rollback
# List available snapshots
python3 main.py --list-snapshots
# Restore from snapshot
python3 main.py --rollback 2026-04-20_143022
# Purge snapshot after confirming all is well
python3 main.py --purge 2026-04-20_143022
Scan modes
| Mode | What it focuses on |
|---|---|
all | Full scan — CPU, memory, disk, security, startup apps |
dev | Developer tools: node_modules, build artifacts, IDE processes |
vulnerability | Security audit: exposed files, open ports, SIP, FileVault |
ui | Visual dashboard only, no cleanup suggestions |
Architecture
main.py ← entry point (argparse, wires all modules)
├── router/
│ ├── skill_router.py ← selects which collectors run per mode
│ └── profiles/ ← JSON config per mode (all / dev / ui / vulnerability)
├── collectors/
│ ├── cpu_memory.py ← CPU %, memory, swap, load average
│ ├── disk_storage.py ← disk usage, large files, old artifacts
│ ├── startup_apps.py ← launch agents, login items
│ └── security.py ← SIP status, FileVault, open ports, file permissions
├── analyzer/
│ └── claude_analyzer.py ← sends scan JSON to Claude, returns scored suggestions
├── dashboard/
│ └── generator.py ← Jinja2 → docs/index.html (GitHub Pages)
├── agents/
│ ├── monitoring_agent.py ← polls metrics continuously, fires alerts
│ ├── remediation_agent.py ← applies safe fixes with rollback support
│ ├── report_agent.py ← generates markdown/HTML reports
│ └── scheduled_scan_agent.py ← wraps main scan for cron/CI
├── rollback/
│ └── rollback_agent.py ← snapshot, restore, purge
├── post_check/
│ └── health_checker.py ← post-cleanup verification (6 checks)
├── data/ ← scan output JSON
└── docs/ ← generated dashboard HTML (served via GitHub Pages)
GitHub Pages setup
- Push this repo to GitHub
- Go to Settings → Pages → Source and set it to the
docs/folder - Add
ANTHROPIC_API_KEYto Settings → Secrets → Actions - The workflow at
.github/workflows/scan.ymlruns a full scan daily at 8 AM UTC
Environment variables
| Variable | Default | Purpose |
|---|---|---|
ANTHROPIC_API_KEY | required | Claude API access |
AI_OPS_MODEL | claude-sonnet-4-6 | Claude model to use |
AI_OPS_MODE | all | Default scan mode |
AI_OPS_OUTPUT_DIR | docs | HTML dashboard output directory |
AI_OPS_DATA_DIR | data | Scan JSON output directory |
Repository structure
.
├── main.py ← AI Ops entry point
├── requirements.txt
├── agents/ ← monitoring, remediation, report, scheduled scan
├── analyzer/ ← Claude AI integration
├── collectors/ ← system metric collectors
├── dashboard/ ← HTML dashboard generator
├── post_check/ ← post-cleanup health verifier
├── rollback/ ← snapshot & restore
├── router/ ← mode-based collector routing
├── data/ ← scan output data
├── docs/ ← generated dashboard (GitHub Pages)
├── python-learning/ ← weekly Python + AI learning projects
│ ├── week2_servicenow_fetcher.py
│ ├── week3_ticket_classifier.py
│ ├── week5_agent.py
│ └── README.md
├── otto/ ← SaaS marketing web mockup
│ ├── index.html
│ └── README.md
└── design-grill-extracted/ ← Clean Architecture + DDD reference notes
Security
Secrets are kept in .env (gitignored). The .env.example file shows all required variable names with no values. Never commit .env or .mcp.json.