wet-compress

Compress Claude Code context by intercepting tool results and reducing token usage.

39|6|Updated Mar 14, 2026

Search Tags:#performance #context management #claude #compression #agent optimization #tool results

Ready to run

npx skills add https://github.com/buildoak/wet --skill wet-compress

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: wet-compress
Source: https://github.com/buildoak/wet/tree/main/skill
Command: npx skills add https://github.com/buildoak/wet --skill wet-compress

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill addresses the issue of Claude's context window becoming bloated with stale or redundant tool results, leading to performance degradation and potential auto-compaction.

Core Features & Use Cases

Context Profiling: Analyzes token distribution within the context to identify compressible items.
Selective Compression: Applies deterministic compression to CLI outputs and LLM-based rewriting for agent returns and search results.
User-Guided Optimization: Presents a clear compression plan for user approval before execution.
Use Case: When working on a long-running agent task, this Skill can surgically remove old git status outputs and summarized agent returns, keeping the active context lean and Claude's reasoning sharp.

Quick Start

Use the wet-compress skill to analyze and compress the current context.

Frequently Asked Questions about wet-compress

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I prevent Claude context window bloat from stale tool results?▼

Context profiling analyzes token distribution within the active Claude session to identify compressible items, presenting a clear optimization plan for user approval before selectively applying deterministic or LLM-based compression to tool results.

How to compress old agent returns and CLI outputs in a long-running Claude session?▼

Yes, you can manage context health and prevent auto-compaction by intercepting tool results and applying intelligent compression. This meta-compression layer surgically removes redundant data, keeping the active context lean and preventing automatic compaction from triggering.

How does intelligent context compression work for Claude Code agent optimization?▼

Before applying context compression, you need an active Claude Code session with accumulated tool results. The profiling phase analyzes the existing token distribution to identify stale or redundant items, generating a user-approved plan for selective data reduction.

Do I need user approval before compressing Claude Code tool results?▼