dsh-workspace-drift
by Npaobi
项目环境体检助手
安装
dsh plugin --profile web add github:Npaobi/dsh-workspace-driftGitHub 源码安装:首次需按提示配置 allowBuilds 构建授权后重试
安装与环境配置指引、插件开发教程见 DSH 中文社区文档 ↗
安装即在你的机器上以你的权限运行第三方代码——它可读写文件、使用凭据、访问网络,DSH 的工具审批不会为插件代码加沙箱。「检测到 manifest」仅代表发现 dsh.bundle / dsh.plugin 清单,不构成兼容性或安全审查;安装前请审阅源码,不熟悉的插件先在不含密钥的环境试用。
README
目录
dsh-workspace-drift is a DeepSeek Harness plugin for finding the environment differences that stop a project from running after it moves to another computer.
It records a project baseline, checks the current machine, and explains differences in plain language. It never saves the values of API keys, passwords, tokens, or .env files.
Install
Install the plugin into a DSH profile:
dsh plugin --profile web add dsh-workspace-drift
For a local checkout during development:
dsh plugin --profile web add .
Use
Ask the agent to use one of these tools with the absolute path to your project.
Record a baseline
environment_snapshot records the environment that currently works and writes it to:
<workspace>/.dsh/environment-baseline.json
Example request:
Record the development environment for E:\Project\my-app.
Check the current machine
environment_check inspects the workspace without changing files. It checks project type, installed runtimes, package manager, dependency-ready markers, common start scripts, and required configuration names.
Example request:
Check whether E:\Project\my-app has the environment needed to run.
Compare with the saved baseline
environment_compare compares the current workspace with the saved baseline and also performs the current-environment checks. Each finding states what differs, why that may block the project, and the next command or action to take.
Example request:
Compare E:\Project\my-app with its saved environment baseline.
What is collected
- operating-system identifier;
- Node.js, Python, Rust, and Go versions when those tools are installed;
- project types detected from
package.json, Python project files,Cargo.toml, andgo.mod; - whether Node dependencies, Python virtual environments, Rust build output, and Go module checksums are present;
- detected package manager and project lockfiles;
- names of package scripts;
- Git branch and commit identifier;
- names from
.env.example, without their values; - whether required names are set in the current process environment or local
.envfile, without retaining their values.
Privacy and limits
The plugin never stores or returns .env values, API keys, passwords, access tokens, private keys, or cookies. During a check it reads local .env assignments only long enough to determine whether each required name has a non-empty value. It only reads files inside the selected workspace. The first release reads package.json, .env.example, .env, and known lockfile names; it does not install dependencies or modify project files other than its own baseline.
Baseline format
Baselines use JSON format version 1. They contain the workspace path, collection time, plugin version, normalized environment facts, dependency-ready markers, and required configuration names, never secret values. The file is written through a temporary file and rename so an interrupted write does not replace the previous baseline with partial JSON. A malformed or unsupported baseline is rejected rather than silently treated as a match.
What a result means
An empty finding list means the plugin found no critical difference among the facts it can inspect. It does not prove that the project will start, build, connect to a service, or pass tests. The plugin never runs project commands or installs dependencies during a check.
Development
npm install
npm run typecheck
npm test
npm run build
The development rules and privacy contract are in AGENTS.md.
Known limitations and deferred work
- Python, Rust, and Go support checks only basic project markers and dependency readiness. It does not parse every ecosystem-specific dependency manager or prove a program can build.
- A matching baseline is evidence of matching recorded facts, not proof that an application will run successfully.
原始 README: https://github.com/Npaobi/dsh-workspace-drift/blob/master/README.md ↗
同类插件
查看全部 →
dsh-anchored-standard
两阶段 DeepSeek Harness 预设:先 Minimal 对齐的 bootstrap,再切完整 Standard 工具(Project2 98/99)

PicGo-Core
极致的图片上传引擎,CLI 与 API 双支持

awesome-deepseek-harness
DeepSeek Harness(DSH)及其优秀社区插件的精选指南。

awesome-deepseek-harness
DeepSeek Harness (DSH)生态系统:来自dsh-external/hub和公共dsh-plugin主题的精选插件、工具和基础设施。

AI-Novel-Writer
本地优先 AI 小说创作工作台,提供 Windows/macOS 桌面版与 DeepSeek Harness 插件开发预览,支持角色、大纲、章节蓝图、审稿修稿和本地模型。

mcp-for-stata
MCP-for-Stata:把 Stata 集成进你的 agent