token-saver

What This Skill Does

Token Saver v3 is an essential optimization engine designed to minimize API costs and improve response latency for OpenClaw users. By intelligently managing the volume of workspace context—including critical documentation like SOUL.md, USER.md, and MEMORY.md—this skill prevents unnecessary token bloat. Unlike static alternatives, Token Saver v3 is fully model-aware. It cross-references your current active AI model against a comprehensive internal registry of over 24 models (including Claude 3.5, GPT-4o, and Gemini 2.0/3.0) to dynamically calculate context window thresholds. It employs smart file-level compression that respects the intent of specific documentation types: keeping personality-driven files like SOUL.md lightly compressed while applying heavy data-density algorithms to MEMORY.md. This ensures you only pay for the context that provides real utility to the LLM.

Installation

You can install this skill directly via the OpenClaw ecosystem by running the following command in your terminal: clawhub install openclaw/skills/skills/rubenaquispe/token-saver. Once installed, the skill automatically integrates with your active session, enabling persistent model detection and monitoring.

Use Cases

This skill is perfect for long-running AI sessions where large project codebases are being tracked. It is specifically useful for power users working with high-context models like Gemini 3 Pro (2M window), as it helps keep the relevant context within the 'sweet spot' for model recall. It also serves as a critical cost-control tool for developers using premium models like Claude 3.5 Sonnet, where every token counts toward the monthly billing threshold. Additionally, if you find your AI agent 'forgetting' instructions due to context window saturation, Token Saver can clean up obsolete workspace information, creating room for high-priority logic.

Example Prompts

1. '/optimize' — Run this to view the interactive dashboard and check your current context consumption vs. your model's limits.
2. '/optimize compaction balanced' — Apply the standard 60% capacity constraint to ensure optimal performance without sacrificing memory depth.
3. '/optimize tokens' — Manually trigger a workspace sweep to compress current markdown files and save on next-message costs.

Tips & Limitations

• Backups: Token Saver v3 automatically creates backups before compression. If you feel a document was over-compressed, simply run '/optimize revert' to restore your original project structure.
• Custom Thresholds: If you are debugging complex logic, you can bypass dynamic presets by using '/optimize compaction 120' to explicitly set a high-token threshold.
• Limitations: This skill is currently optimized for markdown-based project structures. Binary files or non-standard configuration files may be ignored during the compression sweep.