How to use the textworld.agents.RandomCommandAgent function in textworld

options.nb_rooms = world_size
            options.nb_objects = nb_objects
            options.quest_length = quest_length
            options.quest_breadth = quest_breadth
            options.seeds = game_seed
            game_file, game = textworld.make(options, path=tmpdir)

            # Solve the game using WalkthroughAgent.
            agent = textworld.agents.WalkthroughAgent()
            textworld.play(game_file, agent=agent, silent=True)

            # Play the game using RandomAgent and make sure we can always finish the
            # game by following the winning policy.
            env = textworld.start(game_file)

            agent = textworld.agents.RandomCommandAgent()
            agent.reset(env)
            env.compute_intermediate_reward()

            env.seed(4321)
            game_state = env.reset()

            max_steps = 100
            reward = 0
            done = False
            for step in range(max_steps):
                command = agent.act(game_state, reward, done)
                game_state, reward, done = env.step(command)

                if done:
                    msg = "Finished before playing `max_steps` steps because of command '{}'.".format(command)
                    if game_state.has_won:

def test_game_random_agent(self):
        env = textworld.start(self.game_file)
        agent = textworld.agents.RandomCommandAgent()
        agent.reset(env)
        game_state = env.reset()

        reward = 0
        done = False
        for _ in range(5):
            command = agent.act(game_state, reward, done)
            game_state, reward, done = env.step(command)

options.chaining.max_depth = quest_depth
            options.chaining.max_breadth = quest_breadth
            options.seeds = game_seed
            game_file, game = textworld.make(options)

            # Solve the game using WalkthroughAgent.
            agent = textworld.agents.WalkthroughAgent()
            textworld.play(game_file, agent=agent, silent=True)

            # Play the game using RandomAgent and make sure we can always finish the
            # game by following the winning policy.
            env = textworld.start(game_file)
            env.infos.policy_commands = True
            env.infos.game = True

            agent = textworld.agents.RandomCommandAgent()
            agent.reset(env)

            env.seed(4321)
            game_state = env.reset()

            max_steps = 100
            reward = 0
            done = False
            for step in range(max_steps):
                command = agent.act(game_state, reward, done)
                game_state, reward, done = env.step(command)

                if done:
                    assert game_state._winning_policy is None
                    game_state, reward, done = env.reset(), 0, False

def benchmark(game_file, args):
    infos = textworld.EnvInfos()
    if args.activate_state_tracking or args.mode == "random-cmd":
        infos.admissible_commands = True

    if args.compute_intermediate_reward:
        infos.intermediate_reward = True

    env = textworld.start(game_file, infos)
    print("Using {}".format(env))

    if args.mode == "random":
        agent = textworld.agents.NaiveAgent()
    elif args.mode == "random-cmd":
        agent = textworld.agents.RandomCommandAgent(seed=args.agent_seed)
    elif args.mode == "walkthrough":
        agent = textworld.agents.WalkthroughAgent()

    agent.reset(env)
    game_state = env.reset()

    if args.verbose:
        env.render()

    reward = 0
    done = False
    nb_resets = 1
    start_time = time.time()
    for _ in range(args.max_steps):
        command = agent.act(game_state, reward, done)
        game_state, reward, done = env.step(command)

def make_agent(args):
    if args.mode == "random":
        agent = textworld.agents.NaiveAgent()
    elif args.mode == "random-cmd":
        agent = textworld.agents.RandomCommandAgent()
    elif args.mode == "human":
        agent = textworld.agents.HumanAgent(autocompletion=args.hints, walkthrough=args.hints)
    elif args.mode == 'walkthrough':
        agent = textworld.agents.WalkthroughAgent()
    else:
        raise ValueError("Unknown agent: {}".format(args.mode))

    return agent