DEVELOPER TOOLS/开发者工具·CODING-AGENT
ojuschugh1/sqz
压缩AI编程助手收到的工具输出，省token省成本。Compresses tool output for AI coding agents, deduplicating repeated content into compact references to save tokens and costs.
怎么装：Install: cargo install sqz-cli sqz-mcp # or: brew install ojuschugh1/sqz/sqz · npm i -g sqz-cli · pipx install sqz
02 / 现在的位置02 / Why now
首次发现FIRST SEEN
03 OCT 202603 OCT 2026
增长GROWTH
+0.9%+0.9%
637 → 643 stars · 7 个快照637 → 643 stars · 7 snapshots
讨论DISCUSSION
1 SOURCES1 SOURCES
1 条外部提及1 external mentions
状态STATUS
持续活跃Active
项目已有643个star和43个fork，社区关注度高。Saves an average of 24.7% tokens per compression (178,442 tokens saved in one week).
HN_STORY · 1 points · 1 comments
03 / 它能帮你做什么03 / What it helps you do
Compress tool output before it reaches the model
Deduplicate repeated content into compact references
Save an average of 25% on token usage
View token savings with `sqz gain`
Initialize hooks for multiple coding tools with `sqz init`
04 / 安装04 / Install
$ README 里给出的安装方式。The installation method given in the README.
$ README 里给出的安装方式。The installation method given in the README.
两种装法选一个即可 —— 上面那条给命令行用户，下面那条给写代码的用户。Pick one of the two — the first is for command-line users, the second for people writing code.
使用 curl -fsSL https://raw.githubusercontent.com/ojuschugh1/sqz/main/install.sh | sh 安装。Install with a single command using cargo, npm, Homebrew, or a curl pipe—prebuilt binaries work without a compiler.
05 / 限制与风险05 / Limits and risk
风险不是警告，是可信度的一部分。以下结论只基于文档静态扫描，我们不会执行项目里的任何代码。Risk here is evidence, not an alarm. These findings come from static scanning of the docs; we never execute a project’s code.
限制LIMITS
存在风险Risk found
静态扫描STATIC SCAN
01执行远程脚本Runs a remote scriptreadme
curl -fsSL https://raw.githubusercontent.com/ojuschugh1/sqz/main/install.sh | sh
安装脚本会直接执行远程下载的代码（curl 管道到 sh），这是高风险做法：你等于把电脑的完全控制权交给了 GitHub 上这个仓库的作者，仓库被黑或作者改脚本都会让你中招。建议先下载脚本检查内容再执行，或者用 npm/brew 安装。The curl pipe installation method downloads and runs a script from GitHub, which is a common security concern. Consider using cargo or npm install instead if you prefer not to execute remote scripts.
这里只做静态扫描：读 README、SKILL.md 和依赖清单，不执行代码。没有命中不代表安全。This is static scanning only: we read the README, SKILL.md and dependency list, and never execute code. No findings does not mean safe.
{
"agents": [
{
"agent": "claude-code",
"evidence": "tree: .claude/"
},
{
"agent": "cline",
"evidence": "tree: .clinerules/"
},
{
"agent": "gemini-cli",
"evidence": "tree: .gemini/"
},
{
"agent": "windsurf",
"evidence": "tree: .windsurf/"
},
{
"agent": "claude-ai",
"evidence": "readme: …Browser extension | ChatGPT, Claude.ai, Gemini, Grok, Perplexity | |…"
},
{
"agent": "codex",
"evidence": "readme: …transparent) | `sqz init` | | Codex CLI | AGENTS.md guidance + MCP se…"
},
{
"agent": "vscode",
"evidence": "readme: …io-code&logoColor=white\" alt=\"VS Code\"></a> <a href=\"https://addo…"
},
{
"agent": "opencode",
"evidence": "readme: …itten: ```sh sqz init --only opencode # just OpenCode,…"
},
{
"agent": "copilot",
"evidence": "readme: …te .github/hooks/sqz.json for Copilot agent CI runs sqz compress <text>…"
}
],
"apiKey": null,
"docker": null,
"taxonomy": {
"scores": {
"design": 0,
"ai-tools": 6,
"dev-tools": 8,
"indie-web": 0,
"skill-agent": 13,
"productivity": 6
},
"primary": "dev-tools",
"secondary": "coding-agent"
},
"localRuntime": null,
"skillMdTotal": 0,
"skillMdErrors": [],
"treeTruncated": false,
"categoryScores": {
"data": 0,
"agent": 3,
"media": 0,
"design": 0,
"browser": 0,
"devtool": 1,
"security": 0,
"marketing": 0,
"productivity": 0
},
"classification": [
{
"signal": "README ships an mcpServers config for this server",
"weight": 0.7
},
{
"signal": "MCP topic + README",
"weight": 0.6
},
{
"signal": "description mentions AI",
"weight": 0.2
}
],
"scannedSources": [
"readme"
],
"skillMdFetched": 0
}