data-analyst-aiws

AI Workspace

This repository is a small AI workspace for Claude Code.

It is built around one idea: useful Claude workflows should be packaged as installable plugins, not left as personal setup, scattered prompts, or one-off local habits.

The platform currently gives teams three reusable layers:

• core-aiws for shared process and improvement workflows
• memory-aiws for shared cross-project memory contracts
• data-analyst-aiws for analyst-specific workflows

What It Is For

Use this platform when you want Claude Code to behave less like a blank assistant and more like a reusable working system.

It is meant for teams that want:

• shared operating procedures across Claude sessions
• reusable domain workflows instead of ad hoc prompting
• memory boundaries between project memory, shared memory, and runtime state
• a path to ship more domain plugins over time

In practice, this means you can install a common foundation once, then add domain plugins on top of it.

How The Platform Is Structured

The platform is intentionally split into composable plugins rather than one large monolith.

core-aiws

The shared process layer.

It provides:

• the platform SOP
• the public /aiws-improve workflow
• shared protocols that other plugins can depend on

memory-aiws

The shared memory layer.

It defines:

• the cross-plugin shared-memory model
• import and export contracts
• automatic candidate capture and consolidation rules
• the boundary between authoritative docs and advisory shared memory

data-analyst-aiws

The first real domain plugin.

It currently provides:

• data-analyst-forecast

This plugin is intentionally primed, not blank-slate: it ships domain references and workflow structure so users get useful behavior on day one.

Why This Is Extensible

This repository is not only an analyst plugin repo. It is a platform for adding more plugins with the same architecture.

The extensibility model is:

• core-aiws stays the shared process foundation
• memory-aiws stays the shared memory foundation
• each new domain plugin adds its own skills, agents, references, and bootstrap docs

That means future plugins can follow the same pattern without reinventing the platform:

• lawyer-aiws
• marketologist-aiws
• product-manager-aiws
• other domain-specific plugins

A new plugin should not need to reimplement:

• SOP
• self-improvement workflow
• shared memory contracts
• project-memory boundaries

It should only contribute:

• domain workflows
• domain agents
• domain references
• domain-specific bootstrap and integration guidance

Install

In Claude Code:

/plugin marketplace add sashakang/ai-workspace
/plugin install core-aiws@ai-workspace
/plugin install memory-aiws@ai-workspace
/plugin install data-analyst-aiws@ai-workspace

Then restart Claude Code if prompted.

If you already installed the marketplace earlier and want the latest plugin state, refresh and reinstall the relevant plugin.

Current State

This platform is installable now as an early alpha.

What is real today:

• shared process foundation
• shared memory contract layer
• one example domain plugin for data analysts

What that means in practice:

• the architecture is real
• the install path is real
• the analyst workflows are real
• the platform is still early and intended to expand with more plugins over time

Repository Layout

ai-workspace/
├── core-aiws/
├── memory-aiws/
├── data-analyst-aiws/
└── docs/

Each plugin is independently installable from the marketplace, but they are developed together because they share contracts and architecture.

Development

Repository path:

~/Documents/ai-workspace/

Local runtime testing:

claude \
  --plugin-dir ~/Documents/ai-workspace/core-aiws \
  --plugin-dir ~/Documents/ai-workspace/memory-aiws \
  --plugin-dir ~/Documents/ai-workspace/data-analyst-aiws

End users should install through the marketplace, not by cloning or symlinking the repo.

Read More

• Platform architecture
• GitHub and rollout strategy