After setup on Windows

Some commands here show a name in angle brackets until you fill it in. Add your names on the setup page, and every command on every page fills them in.

What zsh through wezterm give you

If you left the rows from zsh through wezterm ticked:

WezTerm with a Rose Pine theme. Ctrl+V pastes, Ctrl+C copies when text is selected, and Shift+Enter starts a new line in a Claude message.
zsh with a starship prompt, grey suggestions from your history (the right arrow accepts one), syntax highlighting, and Ctrl+R to search everything you have typed.
z <part of a folder name> jumps to a folder you have visited; y opens the yazi file browser (q quits); rg, fd, bat, eza (as ls, ll, la), and btop.
A Claude Code status line with the model, how full the conversation is, and the git branch.
Through mise: Node with pnpm, Python with uv, and Rust. A project can pin its own versions in a mise.toml.
A browser Claude Code can drive inside Ubuntu, through Playwright, for testing apps and reading pages.

The job-search workspace

If you ticked job search: resume, LinkedIn, and outreach drafting with a check for writing that sounds like AI, and a local page for tracking roles, applications, and contacts. It is a folder at ~/job-search, backed up to a new private GitHub repo named job-search. To turn a resume into a PDF it uses the Chrome from step 3, which you installed yourself; the installer does not add a browser. If you skipped it, run ./install.sh again and tick job search.

Start a session there

Ubuntu

cd ~/job-search && claude

Ask Claude:

Ask Claude

Interview me and fill in my-voice.md.

Then:

Ask Claude

Help me write my evidence files from my old resume at /mnt/c/Users/<your Windows name>/Downloads/resume.pdf.

A Word resume reads best saved as a PDF first.

Bring over Claude Desktop work

In Claude Desktop: Settings, Privacy, Export data. Download the file from the email and rename it claude-export.zip in Downloads.
Unpack it outside the vault, so the whole chat history stays out of the backups. Ask Claude "Unpack claude-export.zip from my Downloads into ~/claude-export", or by hand:
Ubuntu
```
unzip "/mnt/c/Users/<your Windows name>/Downloads/claude-export.zip" -d ~/claude-export
```

cd ~/memory && claude, then ask Claude:

Ask Claude

My Claude Desktop export is in ~/claude-export. Read it and write what should carry over into Auto Memory: facts about me, my work, the people I mention, and decisions I made. One fact per file, each with the date of the chat it came from. Skip anything that only mattered in that chat, then list what you wrote.

Then /memory-groom moves what lasts into the curated notes.

If you added the job-search workspace: cd ~/job-search && claude, then ask Claude:

Ask Claude

My Claude Desktop export is in ~/claude-export. Find my job-search chats and projects. Put facts about my work into evidence/ with where each one came from, and anything I only said from memory into evidence/from-memory.md. Put my earlier drafts into a drafts folder, and use any instructions I gave about my writing to fill in my-voice.md.

Ask Claude to review a draft before reusing it.

Delete ~/claude-export and the export zip when you are done.

Use the vault from Claude Desktop

In Claude Desktop, Settings, Extensions: install Filesystem from Anthropic and give it C:\Users\<your Windows name>\Documents\Memory. Create a project called Memory with these instructions:

Text

My memory vault is the folder C:\Users\<your Windows name>\Documents\Memory, opened through the Filesystem extension.
At the start of every chat, read Bearing.md. Read Atlas.md when you need to know more about me.
Write new facts only to the Auto Memory folder: one fact per file, with today's date in it.
Never edit, move, or delete any other file in the vault.

Desktop runs none of the hooks or skills and cannot push to GitHub. The next /bod or /memory-groom in Ubuntu picks up what it wrote and backs it up.