Standalone skill test runner for AI agent skills. SkillBench runs *.skillbench.{ts,js,mts,mjs} suites in isolated workspaces and writes deterministic reports.
npx @swarmclawai/skillbench@latest init --write
npx @swarmclawai/skillbench@latest run
npx @swarmclawai/skillbench@latest reportimport { defineSkillSuite, fileContains, exitCode } from "@swarmclawai/skillbench";
export default defineSkillSuite({
name: "demo-skill",
cases: [
{
name: "writes a note",
workspace: { "input.txt": "hello\n" },
command: ["node", "agent.mjs"],
assertions: [exitCode(0), fileContains("note.md", "hello")]
}
]
});| Command | Purpose |
|---|---|
skillbench init |
Print or write a sample suite |
skillbench list [path] |
List discovered suites and cases |
skillbench run [path] |
Run suites and write .skillbench/latest.json |
skillbench report [path] |
Render a Markdown report |
skillbench help-agents |
Print the machine-readable command catalog |
Every data-returning command supports --json.