Skills
Agent-readable instructions for working with moxzi — the same files we use in-house. Each is a single markdown file, served raw, so you can point Claude Code, Cursor or any agent that reads skills straight at the URL and it will know the commands, the gotchas and the failure modes.
# Claude Code
mkdir -p .claude/skills/moxzi-cli
curl -o .claude/skills/moxzi-cli/SKILL.md https://moxzi.ai/skills/moxzi-cli.mdStart with the router skill if you want one file that decides which of the others applies.
moxzi — the router
Decides which sub-skill a request needs: local compilation, server hosting, browser hosting or trustless on-chain builds. Read this first and fetch only what the task needs.
moxzi-cli
Compiling Motoko locally: the build, link, deps, info and cache commands, package
resolution through mops or --package, the artifacts produced, and how to
read a diagnostic.
moxzid-server
Running and operating the server: the manifest, every HTTP endpoint, bearer-token auth, readiness, durability and backup, resource limits, and upgrade safety.
moxzi-web
Actors in a browser: page versus worker transport, install and upgrade, IndexedDB persistence, deadlines and outcalls, the COOP/COEP requirement, and the bundler configuration that trips everyone up.
moxzi-onchain-builds
Builds anyone can verify: driving a paid build, the escrow model end to end, CLI identity and funding, operating a compiler canister, and the source-privacy trade-off.