/ci

<input> <check_type>$ARGUMENTS</check_type> </input> <usage>

Command	Effect
/crew:ci	Run all checks
/crew:ci test	Tests only
/crew:ci lint	Linting only
/crew:ci format	Format check only
/crew:ci typecheck	Type checking only

</usage> <process>

Use the Bash subagent for CI checks. It runs in a separate context window, preventing output pollution in the main thread.

const pm = Bash({
  command:
    "[ -f bun.lock ] && echo bun || ([ -f pnpm-lock.yaml ] && echo pnpm || echo npm)",
}).trim();

const commands = {
  test: `${pm} run test || ${pm} exec vitest run`,
  lint: `${pm} run lint || ${pm} exec biome lint .`,
  format: `${pm} run format:check || ${pm} exec biome format . --check`,
  typecheck: `${pm} run typecheck || ${pm} exec tsc --noEmit`,
  all: `${pm} run ci || (${pm} run lint && ${pm} run test && ${pm} run typecheck)`,
};

Task({
  subagent_type: "Bash",
  prompt: `Run CI check: ${checkType}

1. Run: ${commands[checkType]}
2. Report results:
   - Pass: "ALL CHECKS PASSING"
   - Fail: "[ERROR|WARN] type: file:line - message" (1 line per issue)

The Bash subagent handles large output in its own context - no temp files needed.`,
  description: `ci-${checkType}`,
  run_in_background: true,
});

</process>