Skill Detail

Compress repeated agent context and command output before it reaches coding agents with sqz

<p>Reduces token burn by compressing command output and deduplicating repeated file reads before they are sent to Claude Code, Codex, Cursor, and related coding-agent tools.</p>

Developer ToolsMulti-Framework
Developer Tools Multi-Framework Published
⭐ 165 GitHub stars
INSTALL WITH ANY AGENT
npx skills add agentskillexchange/skills --skill compress-repeated-agent-context-and-command-output-before-it-reaches-coding-agents-with-sqz Copy
Works best when you want a reusable capability, not another fragile one-off prompt.
At a glance
Tools required
sqz CLI plus a supported coding-agent client such as Claude Code, Codex, Cursor, Windsurf, Cline, Gemini CLI, or OpenCode
Install & setup
Install via the project installer script or npm, then run sqz init (optionally with –only or –skip) to add the output-compression hooks to the supported agent clients you use.
Author
Ojus Chugh
Publisher
Open Source Maintainer
Last updated
Apr 27, 2026
Quick brief

Use sqz when an agent is wasting tokens on large command output, repeated file reads, or verbose diffs across an iterative coding session. It is especially useful in repo-scale work where the same context gets pulled multiple times and you want the agent to keep working with less context bloat instead of manually pruning prompts.

How it works

What this skill actually does

This is skill-shaped because the workflow is narrow and operational: install the hook, let sqz intercept tool output, and shrink or deduplicate what the agent receives. It is not a generic LLM utility catalog entry or a broad productivity app listing. Invoke it when the job is specifically to reduce agent context volume while preserving usable command and file output.