Runtime APIs
SkillFiles & storagePrefer Bun-native APIs, else Node built-ins with the node: prefix. Use when importing a Node built-in (fs, path, child_process, crypto, os, util), reading or writing files, spawning a process, or choosing between a Bun and a Node API.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Runtime APIs skill
What this skill tells your AI
The instructions your AI receives, as published by nrjdalal/zerostarter in .agents/skills/runtime-apis/SKILL.md and read by ahel’s review.
Two rules, in order:
- Bun-first. When the file runs under Bun and a native equivalent exists, use it:
Bun.file,Bun.write,Bun.spawn/Bun.spawnSync,Bun.Glob,Bun.YAML,Bun.serve. - Else the
node:prefix. Reach for the Node built-in with its protocol prefix:import { join } from "node:path",require("node:fs"). Never the bare specifier: writenode:path, notpath.
The prefix is mandatory even in Bun-only code: it marks the import as a built-in rather than an npm package, so resolution is identical under every runtime.
Not everything runs under Bun
Bun.* exists only on the Bun runtime. These areas run on Node in at least one of their environments, so node: is load-bearing and Bun.* is a crash, not a style choice:
| Area | Runtime | Rule |
|---|---|---|
packages/cli/** | Node | node: only. Published to npm, launched via npx (bin shebang #!/usr/bin/env node); Bun.* breaks every npx user. |
web/next/** | Node + Bun | node: only. next dev runs under the system Node locally; Docker (bun server.js) and Vercel (bunVersion, built via bun run --bun) run it under Bun. Stays portable. |
packages/env/** | Node + Bun | node: only. Imported by web (Node and Bun) and api (Bun); stays portable. Stricter for web-next.ts and the package index: client components import them, so they carry no node:* at all, prefixed or not. The .env load (node:path + dotenv) lives in load-dotenv.ts, which only the server targets import; a new server target imports @/load-dotenv first, and a script that reaches env through web-next or the index imports @packages/env/load-dotenv itself. |
.github/workflows/* (actions/github-script) | Node | require("node:..."). |
Bun-first therefore applies to the Bun-only files: .github/scripts/*.ts, packages/scripts/src/*.ts (bun x.ts) and the end-to-end helper tests/stack.ts (bun test). The CLI test files run under bun test but mirror the CLI's node: style on purpose.
Bun equivalents
| Need | Bun (preferred) | Node fallback |
|---|---|---|
| read file text | await Bun.file(p).text() | node:fs readFileSync |
| read file bytes | await Bun.file(p).arrayBuffer() | node:fs readFileSync |
| write file | await Bun.write(p, data) | node:fs writeFileSync |
| file exists | await Bun.file(p).exists() | node:fs existsSync |
| spawn a process | Bun.spawn / Bun.spawnSync | node:child_process |
| glob | new Bun.Glob(pattern) | (none) |
| parse YAML | Bun.YAML.parse | (none) |
| serve HTTP/WS | Bun.serve | @hono/node-server (Vercel only) |
Built-ins with no Bun equivalent (always node:)
node:path, node:os, node:util, node:crypto, node:readline/promises, and the directory/sync parts of node:fs (existsSync in sync code, mkdirSync, mkdtempSync, readdirSync, rmSync). Bun ships no Bun.path/Bun.os, so node: is the idiomatic API for these even in pure-Bun files.
Audit
No bare (unprefixed) built-in specifier may exist anywhere. The $B alternation is the full Node built-in list, so a new bare import cannot slip through; extend it only when Node ships a new module. This prints nothing when clean:
B='assert|async_hooks|buffer|child_process|cluster|console|constants|crypto|dgram|diagnostics_channel|dns|domain|events|fs|http|http2|https|inspector|module|net|os|path|perf_hooks|process|punycode|querystring|readline|repl|stream|string_decoder|sys|timers|tls|trace_events|tty|url|util|v8|vm|wasi|worker_threads|zlib'
rg -n "from ['\"]($B)(/[a-z_]+)?['\"]|require\(['\"]($B)(/[a-z_]+)?['\"]\)" -g '!**/dist/**' -g '!.claude/worktrees/**' --hidden
The full per-file inventory lives in index.md: every file, its runtime, and the node: modules it imports. Regenerate it when the surface changes:
rg -n "node:[a-z/_]+" -g '!**/dist/**' -g '!.claude/worktrees/**' --hidden
Signals
- GitHub stars
- 63
- Forks
- 11
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
runtime-apis- Source
- github.com/nrjdalal/zerostarter