Launch feedback: where did it break?

[dgesteves]

onsystem launched today, together with ondocs and signoff-ui. This is the thread for first impressions of onsystem: where it broke, what confused you, what it got wrong. A short comment is fine.

What helps most, in this order:

1. A false positive. A finding on code that is fine. Paste the snippet, the rule id, and what you expected. The output of npx onsystem inspect shows what it read.
2. A first run that didn't work. npx -y onsystem check . or the Claude Code plugin, in your own project. Say your setup (shadcn/ui, a monorepo, a design-system package, MUI) and your agent or editor.
3. Docs that don't work as written. The page and the step.

If it turns out to be a bug, I'll open an issue and link it here. How I handle support and AI-assisted PRs: SUPPORT.md.

What comes next is in the roadmap. A reaction or a comment on one of its issues moves it up.

Try it without installing anything:

npx -y onsystem check .

In Claude Code:

/plugin marketplace add dgesteves/onsystem
/plugin install onsystem@dgesteves

Docs: https://onsystem.vercel.app/docs · What the benchmark measured, and what it can't show: https://onsystem.vercel.app/benchmark

Trying the other two? ondocs (Ask AI for your docs) and signoff-ui (diff review and tool approvals for agent UIs) have their own launch threads in ondocs Discussions and signoff-ui Discussions.