tanphuc16797/workspace-wiki — explained in plain English
Analysis updated 2026-05-18
Stop Claude Code from leaking rules between client A and client B by binding each repo to its own workspace via .claude/wiki.json.
Share a consistent set of agent prompts and pipeline rules across a team using the shared engine layer.
Layer reusable stack packs (frontend, web APIs, event-driven) on top of a workspace to inject the right patterns.
Use /use-wiki and /update-wiki around each task so the wiki stays current after every coding session.
| tanphuc16797/workspace-wiki | 1tdspg-26/front-aula5-1sem | catowabisabi/heso-ai-orchestrator | |
|---|---|---|---|
| Stars | 18 | 18 | 18 |
| Language | HTML | HTML | HTML |
| Setup difficulty | easy | easy | moderate |
| Complexity | 2/5 | 1/5 | — |
| Audience | developer | vibe coder | developer |
Figures from each repo's GitHub metadata at analysis time.
Only Claude Code is officially supported and the install is a one-line script from the latest GitHub release, with an optional SHA256 verified path.
Workspace Wiki is an add-on for Claude Code, the command-line and editor extension version of Claude. It tries to solve a problem that hits teams who use Claude Code across many different projects or client companies at once: the assistant can start blurring context, pulling rules or names from one codebase into another. This repository sets up a structured knowledge layer that is locked to whichever project folder you are working in. The mental model has two parts. The shared engine holds general agent prompts, pipeline rules, and slash commands. Alongside it sit any number of workspaces, each one a separate sandbox of company or project knowledge. The active workspace is stored per project in a small .claude/wiki.json file, so switching codebases switches the context automatically. Between the engine and the workspace, you can also turn on packs, which are bundles of stack-specific patterns for things like web APIs, event-driven systems, frontend work, or product knowledge. The author lays out five non-negotiable rules. Workspace isolation is required, agents are fed runtime context rather than just static documents, packs act as reusable reasoning modules, only Claude Code is officially supported, and knowledge is prioritised in a fixed order with contracts at the top and general domain knowledge at the bottom. The audience the author has in mind is teams running multiple projects, tech leads who want consistent agent output, and solo builders who want repeatable workflows. Installation is by a one-line curl or PowerShell command that pulls a script from the latest GitHub release, with a separate verified-install path that checks SHA256 hashes first. Daily use is through slash commands inside Claude Code, such as /list-workspaces, /switch-workspace, /use-wiki for starting a task, and /update-wiki after coding. Onboarding guides are published in Vietnamese and English through GitHub Pages. The project is MIT licensed and described as best-effort and may move to maintenance mode if the maintainer's capacity changes.
Add-on for Claude Code that locks project-specific knowledge into per-folder workspaces so the assistant stops mixing rules and names from unrelated codebases.
Mainly HTML. The stack also includes HTML, Bash, PowerShell.
MIT license, so you can use, modify, and redistribute it freely, including in commercial work, as long as you keep the copyright notice.
Setup difficulty is rated easy, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.