Token Optimizer

Estimates current context window usage, identifies the largest consumers, and recommends or applies optimization strategies before the context limit is reached.

Purpose

Analyze the composition of the current conversation context, estimate token usage as a percentage of the model's context window, identify the top sources of token bloat, and recommend specific actions to reduce usage — getting user approval before discarding any content.

Trigger

Apply when:

• User says "token usage", "optimize context", "running out of context", "context window full", "how much context left"
• "Token 用量", "節省 token", "壓縮 context", "context 快滿了"
• Proactively when estimated context usage exceeds 70%
• Before starting a large operation that will add significant tokens (e.g., reading many large files)

Do NOT trigger for:

• Normal conversations with low estimated context usage (below 50%)
• Contexts where no optimization is possible or needed