GitHub - RobustaRush/seedkit: Cook a Django project well — the agent-skill take on cookiecutter

Robusta Seedkit 🌱

An agent skill to start new Django projects or extend existing ones.

/seedkit SaaS landing + waitlist, GDPR-friendly stack (mail, analytics, error reporting), VPS deploy

/seedkit add proper auth — magic link, lockout on brute force, optional 2FA

/seedkit look at our repo and tell us what's worth adding next

Helps you with: Python deps & venvs, settings for dev vs prod, custom user model, social & password login, passwordless magic-link login, brute-force protection, background jobs, async views, WebSockets, Redis caching, S3 for static & media, outbound email, Tailwind without Node, GDPR-safe analytics, security headers, CSP headers, production error tracking, structured logs, N+1 query detection, safe migrations, linting & formatting, type checking, scheduled DB backups, Docker for local dev, auto-HTTPS reverse proxy, CI pipeline — and more.

Install

Claude Code (plugin)

/plugin marketplace add RobustaRush/seedkit
/plugin install seedkit@robusta

Other agents (Cursor, Codex, OpenCode, Gemini CLI, …)

Via the skills CLI — installs into whichever agent dirs it detects:

npx skills add RobustaRush/seedkit            # project scope
npx skills add RobustaRush/seedkit -g         # global (all your projects)
npx skills add RobustaRush/seedkit -a cursor  # pin to one agent

Then, in whatever empty directory you'd like to populate:

Two variants

Two skills ship in this repo.

/seedkit started from the actual package docs — the reference files have exact package versions, config snippets, and known pitfalls.

/seedkit-slim relies only on what the model already knows — no reference files.

Both go through the same test cycle: generate code, boot it, check it. Failures get fixed in the skill text.

What's in the skill

The skill targets the problems that show up in LLM-generated Django code that doesn't run out of the box: stale package versions, outdated patterns, wrong async defaults, missing production wiring. The testcase loop's job is to find those failures and fix them in the instructions, so the next generation doesn't repeat them.

It also saves you the prompt-writing: the instructions that would otherwise need a paragraph per project are baked in.

What you get

Ready-to-run Django project.
Pick your stack — real alternatives at every step.
Smart additions: /seedkit add [feature] safely appends code without overwriting.
No unused boilerplate.
Sane defaults out of the box.

Project Status

This is a fresh project under active development. While the skill is verified against the nine core scenarios in seedkit-examples, we are still mapping out how the agent behaves outside that set.

The testcase harness currently runs only against Claude Sonnet. Other models (Opus, Haiku, GPT, Gemini) are not yet covered — they may work, but skill quality on those models is not verified.

Production deployment scenarios (VPS, Fly, GitHub-SSH) still need verification — they are wired up in the skill but not yet exercised end-to-end against real targets.

If you run into issues, strange behavior, or have ideas for new integrations, please open an issue. Feedback is welcome.

Contributing

The most valuable contribution right now: run the testcase harness against other models (Opus, Haiku, GPT, Gemini) and send the logs back. The skill is only verified on Claude Sonnet today, so cross-model coverage is what the project needs most. Point run-tests.sh at the model you want to exercise and open an issue with the resulting workspace/logs/ output attached.
For proposing changes, please open an issue first. Discuss proposed changes — even one-liners — before submitting a pull request to avoid wasted effort.
Expect long test cycles. Validating agent behavior requires multiple test-review-fix iterations. Each cycle runs all nine end-to-end test cases and takes 1.5–2 hours.

License

_{$ Sorry, you're right. I probably shouldn't have deleted the production database.
Want me to at least write the restore script?}