openclaw-token-optimizer

What This Skill Does

The openclaw-token-optimizer is a specialized audit and configuration suite designed to minimize OpenClaw operational costs and latency. It systematically analyzes your current workspace files, system prompt injections, and memory usage patterns to identify 'token bloat.' By implementing intelligent trimming of files like AGENTS.md and SOUL.md, enabling prompt caching, and configuring context pruning, this skill ensures that your agent spends tokens only on what is strictly necessary. It also assists in configuring heartbeat frequencies and model tiering to shift routine tasks to lower-cost models, significantly reducing your monthly spend without sacrificing agent capability.

Installation

To install this skill, run the following command in your terminal: clawhub install openclaw/skills/skills/ccjingeth/openclaw-token-save

Use Cases

• Reducing Monthly Costs: Ideal for power users who have hit high token usage limits and want to optimize expenditure.
• Speeding Up Sessions: Reduces context size, which directly lowers the time-to-first-token and improves response speed in complex threads.
• Improving Context Relevance: Prevents the model from getting distracted by stale, redundant data in long-running projects.
• Configuring Large Workspaces: Essential for developers managing multi-file projects who need to control which files are injected into the context window.

Example Prompts

1. "I've been spending too much on OpenClaw lately. Can you audit my setup and help me cut down my token usage?"
2. "My sessions feel slow and the context is too large. Can you optimize my openclaw.json to enable caching and trim the memory files?"
3. "Help me configure memory search and context pruning so I stop paying to re-read my entire AGENTS.md and history every turn."

Tips & Limitations

• Incremental Changes: Start by trimming the workspace files before enabling complex features like heartbeat, as file trimming offers the highest immediate ROI.
• Backup Configuration: Always backup your openclaw.json before applying suggested optimizations.
• Model Compatibility: Verify that your AI provider supports prompt caching; if not, the optimizer will skip those settings to prevent configuration errors.
• Estimates: Token savings are estimates based on character counting; results may vary depending on the tokenizer used by your specific LLM provider.
• Maintenance: Set a monthly reminder to re-run the optimization, as workspaces evolve and can accumulate bloat over time.