preporato
v0.1.0
Published
Preporato CLI: pull hands-on task starter kits, self-check locally, and submit for instant rubric-graded feedback.
Maintainers
Readme
preporato
Hands-on tasks from preporato.com, graded without leaving your terminal.
Preporato tasks are build-and-submit projects: you pull a starter kit, do the work on your own machine (usually by driving Claude Code), and submit an evidence bundle that an LLM grades against a public rubric. This CLI runs that whole loop.
Requirements
Node.js 18+. If you use Claude Code, you already have it.
Usage
# one-time: connect this terminal to your account
npx preporato login
# pull a task: starter kit + TASK.md land in ./<task-id>/
npx preporato start cc-codebase-tour
cd cc-codebase-tour
# build. TASK.md is the full brief; your agent can read it too.
claude
# green locally first (when the kit ships a self-check)
npx preporato check
# upload and get the rubric verdict right here
npx preporato submitResubmissions are expected and free to make: fix the flagged criteria and run submit again.
Commands
| Command | What it does |
| --- | --- |
| login | Browser handoff: the login page passes a code to this terminal (or paste it manually) |
| start <task-id> | Download and unpack the task's starter kit, write TASK.md and the task manifest |
| check | Run the kit's local self-check (check.py / check.sh / check.js) |
| submit [file] | Upload the deliverable, wait for grading, print the per-criterion verdict |
| whoami / logout | Session management |
Configuration
Session tokens are stored in ~/.config/preporato/config.json (mode 600). Environment overrides: PREPORATO_API_BASE, PREPORATO_WEB_BASE, PREPORATO_TOKEN (direct access token, for CI).
