token_saver_ClaudeCode
skill
Warn
Health Warn
- License — License: MIT
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 5 GitHub stars
Code Pass
- Code scan — Scanned 4 files during light audit, no dangerous patterns found
Permissions Pass
- Permissions — No dangerous permissions requested
Purpose
This skill provides a diagnostic and optimization tool for Claude Code. It analyzes your setup and applies five layers of configuration changes designed to significantly reduce token consumption during your AI sessions.
Security Assessment
Overall Risk: Low
The automated code scan found no dangerous patterns, hardcoded secrets, or requests for elevated permissions. The tool relies on PowerShell and shell scripts to modify local configuration files (like `settings.json` and `.claudeignore`) rather than executing arbitrary code or making external network requests. Because it interacts directly with your local development environment and configuration, you should review the setup scripts before running them to ensure the automated changes align with your personal workflow preferences.
Quality Assessment
The project is highly active, with its last push occurring just today. It is properly licensed under the permissive MIT standard. The README is comprehensive and well-structured, providing both an audit mode to review current usage and a setup mode to apply changes, alongside detailed documentation on validated impacts. The primary drawback is low community visibility; it currently only has 5 GitHub stars. This means the tool has not been broadly tested or vetted by a large user base.
Verdict
Safe to use, though it is recommended to review the automated setup scripts prior to execution to understand the local configuration changes being made.
This skill provides a diagnostic and optimization tool for Claude Code. It analyzes your setup and applies five layers of configuration changes designed to significantly reduce token consumption during your AI sessions.
Security Assessment
Overall Risk: Low
The automated code scan found no dangerous patterns, hardcoded secrets, or requests for elevated permissions. The tool relies on PowerShell and shell scripts to modify local configuration files (like `settings.json` and `.claudeignore`) rather than executing arbitrary code or making external network requests. Because it interacts directly with your local development environment and configuration, you should review the setup scripts before running them to ensure the automated changes align with your personal workflow preferences.
Quality Assessment
The project is highly active, with its last push occurring just today. It is properly licensed under the permissive MIT standard. The README is comprehensive and well-structured, providing both an audit mode to review current usage and a setup mode to apply changes, alongside detailed documentation on validated impacts. The primary drawback is low community visibility; it currently only has 5 GitHub stars. This means the tool has not been broadly tested or vetted by a large user base.
Verdict
Safe to use, though it is recommended to review the automated setup scripts prior to execution to understand the local configuration changes being made.
Professional token optimization skill for Claude Code - Diagnóstico e otimização de consumo de tokens com 5 camadas de otimização
README.md
token-saver: Otimização Profissional de Tokens para Claude Code
English version below ⬇️
🇧🇷 Português
O que é token-saver?
token-saver é uma ferramenta profissional de diagnóstico e otimização de consumo de tokens no Claude Code. Reduz o consumo de tokens em até 85% através de 5 camadas de otimizações validadas.
⚡ Quick Start (3 passos)
Instale o skill:
- Copie
token-saver.skillpara~/.claude/skills/ - Reinicie o Claude Code
- Copie
Execute o diagnóstico:
/token-saver AUDITAplique as otimizações:
/token-saver SETUP
🎯 5 Camadas de Otimização
| Camada | Otimização | Redução | Impacto |
|---|---|---|---|
| 1 | Setup único (Tool Search, .claudeignore, settings.json) | −85% | 🟢 Alto |
| 2 | CLAUDE.md e Skills management | −30-40% | 🟢 Alto |
| 3 | Session commands (/compact, /context, /clear) | −60% | 🟡 Médio |
| 4 | Subagent e MCP discipline | −80% | 🟢 Alto |
| 5 | Output compression (Caveman mode) | −40% | 🟡 Médio |
📚 Documentação Completa
- SKILL.md — Guia completo com todas as otimizações
- sources.md — Referências validadas e benchmarks
- CHANGELOG.md — Histórico de versões
🛠 Scripts de Setup Automático
- Linux/macOS/WSL:
scripts/setup.sh - Windows (PowerShell):
scripts/setup-windows.ps1
Execute automaticamente para:
- Configurar
~/.claude/settings.json - Criar
.claudeignoreno projeto - Diagnosticar MCPs instalados
- Auditar CLAUDE.md
🔍 Modo AUDIT vs SETUP
AUDIT (Diagnóstico)
/token-saver AUDIT
- Relatório rápido do consumo atual
- Identifica oportunidades de otimização
- Sem fazer mudanças
SETUP (Aplicar Otimizações)
/token-saver SETUP
- Aplica as 5 camadas automaticamente
- Configura settings.json
- Cria .claudeignore
- Pronto para uso imediato
📊 Impactos Validados
- Tool Search: −85% de tokens de definições de ferramentas
- .claudeignore: −30 a −40% por sessão
- Subagent Haiku: −80% custo de subagentes
- MAX_THINKING=10000: −70% thinking tokens
- /compact a 60%: Previne degradação de contexto
🐛 Bugs Conhecidos (Abril 2026)
| Bug | Workaround |
|---|---|
| ENABLE_TOOL_SEARCH não funciona no Desktop via env var | Usar settings.json |
| Auto mode (10% threshold) não dispara | Setar true explicitamente |
| defer_loading + cache_control incompatíveis | Desabilitar ENABLE_TOOL_SEARCH se necessário |
🤝 Contribua
Encontrou um problema ou tem uma sugestão? Abra uma Issue ou uma Discussion.
📄 Licença
MIT License — veja LICENSE para detalhes.
🇬🇧 English
What is token-saver?
token-saver is a professional tool for diagnosing and optimizing token consumption in Claude Code. Reduces token usage by up to 85% through 5 validated optimization layers.
⚡ Quick Start (3 steps)
Install the skill:
- Copy
token-saver.skillto~/.claude/skills/ - Restart Claude Code
- Copy
Run diagnosis:
/token-saver AUDITApply optimizations:
/token-saver SETUP
🎯 5 Optimization Layers
| Layer | Optimization | Reduction | Impact |
|---|---|---|---|
| 1 | Single setup (Tool Search, .claudeignore, settings.json) | −85% | 🟢 High |
| 2 | CLAUDE.md and Skills management | −30-40% | 🟢 High |
| 3 | Session commands (/compact, /context, /clear) | −60% | 🟡 Medium |
| 4 | Subagent and MCP discipline | −80% | 🟢 High |
| 5 | Output compression (Caveman mode) | −40% | 🟡 Medium |
📚 Full Documentation
- SKILL.md — Complete guide with all optimizations
- sources.md — Validated references and benchmarks
- CHANGELOG.md — Version history
🛠 Automatic Setup Scripts
- Linux/macOS/WSL:
scripts/setup.sh - Windows (PowerShell):
scripts/setup-windows.ps1
Execute automatically to:
- Configure
~/.claude/settings.json - Create
.claudeignorein your project - Diagnose installed MCPs
- Audit CLAUDE.md
🔍 AUDIT vs SETUP Mode
AUDIT (Diagnosis)
/token-saver AUDIT
- Quick report of current consumption
- Identifies optimization opportunities
- No changes made
SETUP (Apply Optimizations)
/token-saver SETUP
- Applies all 5 layers automatically
- Configures settings.json
- Creates .claudeignore
- Ready for immediate use
📊 Validated Impacts
- Tool Search: −85% tool definition tokens
- .claudeignore: −30 to −40% per session
- Subagent Haiku: −80% subagent cost
- MAX_THINKING=10000: −70% thinking tokens
- /compact to 60%: Prevents context degradation
🐛 Known Bugs (April 2026)
| Bug | Workaround |
|---|---|
| ENABLE_TOOL_SEARCH doesn't work on Desktop via env var | Use settings.json |
| Auto mode (10% threshold) doesn't trigger | Set true explicitly |
| defer_loading + cache_control incompatible | Disable ENABLE_TOOL_SEARCH if needed |
🤝 Contribute
Found an issue or have a suggestion? Open an Issue or Discussion.
📄 License
MIT License — see LICENSE for details.
Made with ❤️ for Claude Code developers
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found