new-workspace

name	new-workspace
description	Provision a new eval-runner workspace on disk. Use when the user wants to start a new evaluation project — scaffolds evals/, datasets/, results/, and docs/ directories, personalises CLAUDE.md, and (by default) creates a GitHub repo.
disable-model-invocation	true
allowed-tools	Bash(mkdir *), Bash(cp *), Bash(cat *), Bash(git init *), Bash(git add *), Bash(git commit *), Bash(gh repo create *), Bash(gh auth status), Bash(git push *), Read, Write

Provision Eval-Runner Workspace

Creates a new workspace for evaluation work. The plugin's commands (/eval-runner:setup-eval, /eval-runner:create-eval, /eval-runner:run-eval, /eval-runner:publish-eval, /eval-runner:publish-dataset, /eval-runner:document-eval) operate against this scaffold.

Arguments

$ARGUMENTS is parsed as:

• First positional: workspace name (Train-Case — e.g. My-Eval-Project). Required.
• Second positional (optional): target parent path. Defaults to ~/repos/github/my-repos.
• --private (optional): create the GitHub repo as private. Default: public.
• --local-only (optional): skip GitHub repo creation.

Scaffold

<workspace>/
├── CLAUDE.md               # workspace-level context for Claude
├── README.md
├── evals/                  # one subfolder per eval
│   └── .gitkeep
├── datasets/               # eval datasets (raw + processed)
│   └── .gitkeep
├── results/                # run outputs, scored, with metadata
│   └── .gitkeep
├── docs/                   # rationale, findings, writeups
│   └── .gitkeep
└── .gitignore

Procedure

1. Parse arguments. If name missing, ask.
2. Copy template/ from the plugin into <parent>/<name>/.
3. Personalise CLAUDE.md — inject workspace name, date, and a short preamble describing what is being evaluated (ask the user in one line if not obvious).
4. git init, initial commit.
5. Unless --local-only, gh repo create (public by default, private with --private) and push.
6. Report the workspace path and repo URL.