AutoGineer
Generated banner for AI Memory Workspace Setup
Generated logo for AI Memory Workspace Setup

AI Memory Workspace Setup

Starter kit

First committed by Fryingpannn

Published in Fryingpannn/ai-memory-workspace-starter. Publisher page

21 GitHub stars

What it is

A Hermes-friendly, files-first AI memory workspace that survives fresh chats.
— the description by Fryingpannn

AutoGineer overview, from the brain’s own files

ai-memory-workspace-starter provides a structured, agent-driven installation process to establish a persistent memory workspace within a user's existing file system. Designed to be platform-agnostic, it uses a set of templates and setup prompts to scaffold a workspace that survives individual chat sessions. The system focuses on creating a canonical structure including a user profile, a wiki, and project-specific directories, ensuring that AI agents can maintain context across different sessions without relying on proprietary cloud storage.

How it works

Memory model: Markdown vault

AutoGineer overview, from the brain’s own files

The system functions as a temporary setup utility. When triggered by an agent, it reads the setup instructions in the docs folder and uses the prompts in the prompts/setup folder to guide the user through a multi-stage installation. It copies missing files from the templates/workspace folder into the user's designated workspace directory, creating a standard structure of folders like projects, skills, outputs, and wiki. It also initializes a root USER.md file and a fused AGENTS.md file to serve as the primary knowledge entry points for the agent.

Folders
FolderRoleFiles
docsdocs2
promptsagent-instructions9
templatestemplates18

The agent reads

The agent writes

Requirements

Works with: Claude Code, OpenAI Codex, GitHub Copilot, Cursor, Kiro, OpenCode, Amp, Cline, JetBrains Junie, Devin (ex-Windsurf)

Set it up

  1. Download the brain as a ZIP and unzip it where you keep your notes or project: https://mp-staging.autogineer.com/agent-brains/fryingpannn/ai-memory-workspace-starter/download

  2. Claude Code reads CLAUDE.md from this folder automatically.

  3. cd ai-memory-workspace-starter && claude

  4. AutoGineer scanned these files at commit cf26138.

  5. First committed by Fryingpannn. Published in Fryingpannn/ai-memory-workspace-starter. Licence: MIT License.

Usage

AutoGineer overview, from the brain’s own files

  • Scaffolding a new persistent memory workspace for an existing AI agent.
  • Standardizing memory structures across multiple agent-enabled projects.
  • Integrating a portable wiki and user profile into an existing file-based development environment.
“Set up my AI memory workspace from this repository.”
The agent will initiate the setup flow, confirm the workspace location, and scaffold the necessary folders and files.
“Verify my workspace memory.”
The agent will run the acceptance test defined in the workspace to ensure the memory structure is correctly linked and accessible.

Documentation by Fryingpannn

AI Memory Workspace Setup

An agent-run installer for adding portable, files-first memory to a new or existing outer workspace.

This repository is not the user's workspace. The user's existing file-aware agent reads SETUP.md, temporarily downloads this package when needed, and installs only the missing scaffold files from templates/workspace/.

One Setup Prompt

Paste this into the agent you already use:

Set up my AI memory workspace from this repository:
https://github.com/Fryingpannn/ai-memory-workspace-starter

Start with SETUP.md. Use this repository only as temporary setup instructions and a
template source; do not turn its clone into my workspace.

The setup manager asks whether an outer workspace already exists. It uses that folder or creates one after approval. Existing projects stay in place.

What This Package Contains

  • SETUP.md: the only setup entrypoint.

  • prompts/setup/: short, ordered setup stages.

  • prompts/VERIFY-WORKSPACE.md: fresh-session acceptance test.

  • prompts/CONNECT-INNER-PROJECTS.md: optional reviewed bridge for child-project instruction files.

  • templates/workspace/: missing-only files for the user's outer workspace.

  • docs/: reference material for the setup agent and implementing user.

The temporary clone is removed after verification and the final handoff. Its Git history and remote are never copied into the user's workspace.

Setup Flow

  1. Confirm or create the outer workspace.

  2. Copy only missing supporting files from templates/workspace/.

  3. Build the portable root USER.md.

  4. Build the interlinked wiki/.

  5. Analyze approved existing child projects read-only.

  6. Create or review a fused canonical AGENTS.md.

  7. Review or initialize the outer Git repository.

  8. Prove retrieval from a fresh session.

  9. Deliver the optional inner-project bridge prompt, then remove the temporary setup clone.

Ownership Boundaries

The portable workspace owns:

  • root USER.md;

  • the CLAUDE.md compatibility bridge;

  • projects/, skills/, outputs/, and wiki/.

Existing target files are preserved except for a user-confirmed replacement of AGENTS.md with its reviewed fused proposal. Git initialization, commits, remote creation, and pushes remain approval-gated.

This installer uses the file-aware agent the user already has. Installing another harness is a separate optional workflow.

prompts/setup/04-BUILD-WIKI.md

Stage 4: Build the Long-Term Wiki

Read the confirmed workspace purpose from the root USER.md. If the user declined onboarding and the purpose is still unknown, ask only for that one sentence.

Spawn one clean-context wiki architect subagent when supported.

Give it:

  • the confirmed outer workspace root;

  • this prompt;

  • the confirmed workspace purpose from USER.md;

  • Andrej Karpathy's LLM Wiki idea file: https://gist.githubusercontent.com/karpathy/442a6bf555914893e9891c11519de94f/raw/ac46de1ad27f92b28ac95459c782c07f6b8c964a/llm-wiki.md

The source communicates the design pattern. Adapt it to this workspace; do not copy it verbatim.

Required Architecture

Use three layers:

  1. raw sources that remain unchanged;

  2. an incrementally maintained, interlinked Markdown wiki;

  3. a local schema that tells agents how to ingest, query, and maintain it.

Create missing pieces under the outer workspace root:

wiki/
├── index.md
├── SCHEMA.md
├── inbox.md
├── log.md
├── pages/
└── raw/

Preserve existing wiki files. Propose additions rather than replacing them. If a wiki already contains user content, show the adaptation plan and ask before editing it.

Wiki Rules

  • wiki/index.md is the discovery map and must link every durable page.

  • wiki/SCHEMA.md defines page types, links, sourcing, ingest, query, and lint rules.

  • wiki/log.md is append-only and uses dated, consistently formatted entries.

  • wiki/raw/ preserves source records; wiki pages synthesize them.

  • Durable entities and relationships become linked Markdown pages.

  • New evidence may update several related pages and must preserve provenance.

  • Contradictions, uncertainty, and stale claims must be visible.

  • Do not add vector search, embeddings, a database, or RAG.

Seed only confirmed workspace-level knowledge. Project-specific knowledge is added in Stage 5. Link the root USER.md from wiki/index.md; do not duplicate personal facts in a wiki user-profile page. Replace every bracketed placeholder in wiki/pages/workspace-profile.md with the confirmed purpose, scope, and provenance.

Return the created or changed files and the proposed wiki schema for user review.

36 files

Advantages

  • Ensures memory persistence across fresh agent sessions by using a file-based vault structure. evidence: README.md
  • Preserves existing project files by only installing missing scaffold components. evidence: README.md

Limitations

  • Requires an existing file-aware agent to execute the setup process. evidence: README.md
  • The installer repository itself is not a workspace and must be removed after setup. evidence: AGENTS.md

Security

Risk level: No risk found

This brain runs

This brain runs nothing: AutoGineer found no hooks, scripts, plugin code or MCP servers in its files.

Attribution and licence

First committed by
GitHub avatar of FryingpannnFryingpannn
Publisher
Duke Pan (Community)
Source
Fryingpannn/ai-memory-workspace-starter · commit cf26138
Licence
MIT License (files served here)
  • ATTRIBUTION.md (written by AutoGineer)
Synced
Contributors
Fryingpannn

AutoGineer is not affiliated with or endorsed by Fryingpannn.

Download

The ZIP holds the files AutoGineer serves, with ATTRIBUTION.md and the licence. Download ai-memory-workspace-starter.zip