bun-runtime

name	bun-runtime
description	Use Bun for file/script operations — Bun.file, Bun.write, Bun.$ shell, child processes. Do not use Python or Node.js fs for file manipulation, generation, or concatenation.
license	ISC
compatibility	Requires `plaited` CLI and network access
allowed-tools	Bash

Bun Runtime Skill

Two modes

This skill covers both Bun API lookup and Bun-based file/script operations. Use the right tool for the task.

---

Mode 1: API Documentation Lookup

Query the Bun documentation via MCP for API reference questions:

`plaited mcp-client '{"mode":"call-tool","url":"https://bun.com/docs/mcp","tool":"search_bun","args":{"query":"Bun.file API"}}'

[!CAUTION] Common failure modes

1. Invalid JSON from special characters — avoid unescaped backticks (`), unescaped double quotes, or raw newlines inside the JSON string argument. The entire '{}' payload must be valid JSON. Use single quotes around the JSON blob and escape any inner double quotes. Backticks inside the query value will break JSON parsing — omit them or use \``` ``` if absolutely needed.

2. Wrong URL — the correct MCP endpoint is https://bun.com/docs/mcp, not bun.sh/api/mcp. Using bun.sh returns a 404.

Mode 2: File Operations & Scripting (THIS REPO)

Always use Bun for scripting and file operations. Do not use Python, Node.js fs, or shell heredocs for file generation, concatenation, transformation, or any local file scripting task.

When implementing a scripted solution, first search the Bun docs for a native Bun API (Bun.file, Bun.write, Bun.$, Bun.spawn, etc.) rather than reaching for Node.js compat pages (bun.com/docs/runtime/nodejs-compat). For operations where no Bun native API exists (e.g., temporary directories), use the Node.js built-in via os.tmpdir() or os.mkdtempSync() — Bun implements nearly all of node:*.

Why

• AGENTS.md requires: "Prefer Bun over Node.js"
• Using Python or heredocs for file concatenation or generation creates brittle, error-prone workflows (quoting issues, truncation, invisible control characters)
• Bun has native APIs for file I/O that are concise and safe

Available approaches (preferred order)

1. Split into multiple files and re-export

If a single file grows too large for the write tool, split it into logical parts and use a re-export boundary file. This follows the repo's module organization conventions (feature.ts re-exports from parts):

// src/open-responses/open-responses.schemas.ts — all Zod schemas
// src/open-responses/open-responses.ts — re-exports + client implementation

No ad-hoc concatenation needed.

2. Use Bun.write + Bun.file for file concatenation

// Write as: bun run -e "..."
bun run -e '
const parts = [
  Bun.file("part1.ts").text(),
  Bun.file("part2.ts").text(),
];
const combined = await Promise.all(parts);
await Bun.write("output.ts", combined.join("\n"));
'

3. Use Bun.$ for shell pipelines

bun run -e 'await Bun.$`cat part1.ts part2.ts > output.ts`.quiet()'

4. Write a small temp script and run it

bun run /tmp/combine.ts  # where /tmp/combine.ts uses Bun.file/Bun.write

5. Use bun run - for stdin-piped one-liners

For quick scripts without a temp file, pipe code through stdin:

echo 'await Bun.write("output.ts", `hello`);' | bun run -

Also works for reading from files:

bun run - < script.ts

6. Use Bun.spawnSync for running external commands

When you need to run an external CLI tool and capture its output:

bun run -e '
const result = Bun.spawnSync(["git", "log", "--oneline", "-5"]);
console.log(result.stdout.toString());
'

Use Bun.which() to check if a command exists before spawning:

bun run -e '
if (Bun.which("plaited")) {
  const result = Bun.spawnSync(["plaited", "skills"]);
  console.log(result.stdout.toString());
}
'

### What NOT to do

- ❌ `python3 -c "..."` — never use Python for any file scripting or manipulation
- ❌ Heredocs with `cat > file << 'EOF'` — brittle with large content, truncation risk
- ❌ `node -e "fs.writeFileSync(...)"` — Bun API is required
- ❌ External Python scripts (`.py` files) for build steps or code generation — use `Bun.$` (shell) or `Bun.spawn` instead

## When to use Mode 1 (API lookup)

- Looking up Bun runtime APIs (`Bun.file`, `Bun.serve`, `Bun.$`)
- Checking bundler or test runner configuration
- Finding package manager commands or compatibility details
- Verifying Bun-specific behavior vs Node.js

## When to use Mode 2 (File operations)

- Generating, combining, or transforming any source files
- Any local file scripting task
- Concatenating file parts that don't fit the `write` tool's size limits
- Running external CLI tools or checking for executable availability

## See also

- `plaited mcp-client --help` — discover all available MCP operations
- `plaited mcp-client --schema input` — inspect the full input schema
- `AGENTS.md` — section on Bun APIs