rivet-dev rivet
COMMUNITYLABSCO SUMMARY
Ten of them are openspec-* skills — new-change, explore, continue-change, ff-change, apply-change, verify-change, sync-specs, archive-change, bulk-archive-change, and onboard — that walk a contributor through Rivet's own experimental change-management process, from proposing a change through a scripted set of artifacts to implementing, verifying, and archiving it. Three more are narrower jobs: driver-test-runner works file by file through the RivetKit driver test suite across the native and wasm runtimes; sanity-check installs the published rivetkit package into a throwaway project and checks a hello-world actor over HTTP and WebSockets before a release goes out; and stack-merge fast-forwards a whole stack of Graphite-managed pull requests into main, which by its own description needs repository admin rights that bypass branch protection, the Graphite CLI, and the GitHub CLI. That accounts for 13 — the last one carries no description in our data, so we can't say what it does.
This only makes sense for someone already contributing changes to the rivet-dev/rivet monorepo under its OpenSpec process, or cutting a Rivet release. If you came here to build an actor-based backend — the in-memory, auto-persisted, WebSocket-and-queue actors the README spends most of its length on — none of these skills teach that. The README points elsewhere for it: npx skills add rivet-dev/skills, a separate package with its own listing in our data.
READ THE FULL ANALYSIS
What it costs to start. Eleven of the fourteen need a local tool already on the machine — going by what the visible descriptions call for, that reads as git, Graphite's gt, and GitHub's gh — and only two install clean. One more is billed as needing an account or credential; stack-merge's admin-rights requirement is the likely match, though our histogram doesn't tie a class to a specific skill name.
WHAT'S INSIDE
14 showing · 14 totalNothing else to set up — install it and go.
driver-test-runner
Works through Rivet's driver tests one file at a time and keeps a note of what has already passed, so a run that stops can pick up where it left off.
graphite-is-being-a-piece-of-shit-just-merge-my-stack
openspec-apply-change
Works down the task list that planning a change produced, writing the code for each task and ticking it off before moving to the next.
openspec-archive-change
The tidy-up at the end: a finished piece of work is moved into a dated archive folder, and anything left half-done is put in front of you before that happens.
openspec-bulk-archive-change
Retires several finished changes at once, and where two of them edited the same specification, settles it by looking at what the code actually does now.
openspec-continue-change
Writes the next document in a change's plan — the proposal, then the specs, then the design, then the task list — one per run, building on the ones already there.
openspec-explore
A thinking partner for the stage before anything gets built — it reads the code you already have, draws diagrams, and talks options through with you, and writes no code at all.
openspec-ff-change
Writes every planning document a new piece of work needs, back to back, instead of stopping for your review after each one.
openspec-new-change
Opens a folder for a new piece of work and stops at the first blank document, so you can point it in a direction before anything gets written.
openspec-onboard
A guided tour of the OpenSpec way of working: it takes a small job from your own codebase and carries it all the way from first idea to filed away, explaining every step out loud.
openspec-sync-specs
Folds what one change decided into the project's standing specification — a paragraph added here, one reworded there — without declaring the change finished.
openspec-verify-change
Before a piece of work is signed off, everything it promised gets checked against what was actually built, with a report of what is missing or has quietly drifted.
sanity-check
Proves a freshly published version of Rivet's toolkit really works, by installing it from scratch into an empty folder, starting the sample app and checking it answers correctly.
stack-merge
Lands a whole chain of stacked pull requests onto the main branch in one push and closes each of them, instead of feeding them through a merge queue one at a time.
HOW TO GET IT
npx skills add rivet-dev/rivetnpx skills add rivet-dev/rivet --skill <name> --full-depthPick the skill name from the Skills tab — each entry there installs independently.