What is session bloat and why is it dangerous?

Session bloat occurs when an agent accumulates too much context without compacting it, leading to higher costs per request and potential context quality decay where the model loses track of important information.

Do I need existing data for the audit to work?

Yes, the skill first runs a scan on your OpenClaw session data (typically the last 30 days) to analyze your actual usage patterns before suggesting optimizations.

What are 'ghost tokens' in OpenClaw?

Ghost tokens are unnecessary or redundant tokens consumed during LLM sessions, such as excessive context headers or inefficient heartbeat checks, that increase costs without adding value.

How does this skill help save money?

It identifies specific instances where high-cost models like Claude 3.5 Sonnet are used for low-complexity tasks and provides exact configuration changes to reduce token consumption.

Token Optimizer

Name: Token Optimizer
Author: alexgreensh

byalexgreensh

•

862

•

分析与监控

Audits OpenClaw agents to identify token waste, reduce context bloat, and lower operational costs.

Token Optimizer is a specialized utility for OpenClaw users designed to streamline LLM usage by detecting inefficient token patterns and 'ghost tokens' that drive up costs. It scans historical session data, audits model selection for low-stakes tasks like heartbeats, and provides actionable configuration fixes to prevent session bloat and context quality decay. By translating technical token metrics into real-world dollar amounts, it empowers developers to maintain high-performance agents while maximizing their ROI through smarter resource allocation.

主要功能

01Step-by-step configuration fixes for OpenClaw agents

02Waste pattern detection with severity-based reporting

03Automated session scanning and historical data collection

04862 GitHub stars

05Context management and session compaction optimization

06Financial impact analysis with estimated monthly savings

使用场景

01Fixing session bloat in long-running autonomous agents

02Reducing monthly API costs by switching models for low-complexity tasks

03Optimizing expensive cron-based agents that run frequently

主要功能

01Step-by-step configuration fixes for OpenClaw agents

02Waste pattern detection with severity-based reporting

03Automated session scanning and historical data collection

04862 GitHub stars

05Context management and session compaction optimization

06Financial impact analysis with estimated monthly savings

使用场景

01Fixing session bloat in long-running autonomous agents

02Reducing monthly API costs by switching models for low-complexity tasks

03Optimizing expensive cron-based agents that run frequently