Confdiff：支持敏感信息脱敏的配置文件语义差异工具

由自主 AI 代理 Esperanza Volkov 构建并维护的 confdiff 是一款面向配置与结构化数据文件的语义差异工具 1。该工具通过解析数据模型而非文本进行对比，仅报告真实的键值变化 1。它支持 JSON、YAML、TOML、INI/.cfg/.conf、.env、Java.properties、CSV/TSV 和 XML 八种格式，可自动检测格式并以内容嗅探作为回退 1。同时，confdiff 提供 --redact 选项，能将密码、令牌、API 密钥等敏感值替换为形如 «redacted:28c19f» 的稳定指纹，避免在 PR 评论、Slack 或 CI 日志中泄露 1。

confdiff 支持 CLI、npm、Docker、GitHub Action 及作为 git diff 驱动等多种使用方式 1。用户可通过 npm install -g confdiff 全局安装或使用 npx confdiff 直接运行，要求 Node.js ≥ 18 1。项目提供 Docker 镜像 ghcr.io/esperanza-volkov/confdiff，支持 :latest 或 v0.10.0 等版本标签；同时提供 GitHub Action esperanza-volkov/confdiff@v1，支持 redact、comment、fail-on-diff 等输入参数 1。其退出码约定为：0 表示无差异，1 表示存在差异，2 表示用法或解析错误 1。该项目采用 MIT 许可证，版权归 Esperanza Volkov 所有 1。

Confdiff is a semantic difference tool designed for configuration and structured data files that parses data models instead of plain text to report only genuine key-value changes 1. The utility supports eight formats, including JSON, YAML, TOML, INI/.cfg/.conf, .env, Java properties, CSV/TSV, and XML, and features automatic format detection with content sniffing as a fallback mechanism 1. The project is built and maintained by an autonomous AI agent named Esperanza Volkov 1. It supports usage via CLI, npm, Docker, GitHub Action, and as a git diff driver 1.

A standout feature of the tool is the --redact option, which replaces sensitive values such as passwords, tokens, and API keys with stable fingerprints like «redacted:28c19f» to prevent accidental exposure in pull request comments, Slack, or continuous integration logs 1. Developers can install confdiff globally via npm install -g confdiff or execute it directly using npx confdiff, requiring Node.js version 18 or higher 1. For containerized and automated workflows, the project provides a Docker image at ghcr.io/esperanza-volkov/confdiff supporting :latest or version tags such as v0.10.0, and a GitHub Action at esperanza-volkov/confdiff@v1 that supports redact, comment, and fail-on-diff inputs 1. The software utilizes standard exit codes, returning 0 for no differences, 1 for detected differences, and 2 for usage or parsing errors, and is distributed under the MIT license with copyright held by Esperanza Volkov 1.