为什么开始做幻兽帕鲁模组：想把游戏改得更顺手

做模组这件事，是从打游戏开始的。

玩着玩着，总会冒出一些很具体的念头：基地边界能不能再往外一点，箱子能不能多放些东西，终端里的帕鲁能不能少点几次就整理好。它们不一定是什么大问题，但反复碰到，还是会想动一动。

我做这些模组的出发点，大多就是这样。先把自己觉得不顺手的地方改一小块，再看改完以后，是不是更接近想玩的样子。

一个改动，只管一件事

现在整理出来的模组里，有些改的是范围和容量：基地半径、箱子格数、分解器的待处理位置。还有一些改的是操作过程：批量送入、跨页选材料，以及按条件筛选后设置收藏。

这些东西看起来都和“少做几次重复操作”有关，但实际修改的并不是同一件事。

比如分解器的待处理位置变多，不等于加工速度变快。箱子创建时的容量扩大，也不代表存档里原有的箱子会自动迁移。页面里把它们写清楚，和把模组做出来一样有必要。不然只看一个“容量翻倍”，很容易以为其他部分也一起变了。

我倾向于把这些改动拆开。需要哪个就装哪个，不必为了一个功能，把其他规则也跟着改掉。

少一点重复，也保留最后的决定

做批量操作时，我在意的还有一件事：省下的是操作次数，最后处理哪些帕鲁，仍然应该由玩家看清楚。

浓缩机快速选材会帮忙翻页、找到符合条件的材料，但不会替你点下最后的浓缩。分解机批量送入则先给出名单，再等一次确认。它们的后果不同，提示也不能完全一样。

收藏筛选也是同样的思路。先选条件，再预览，最后设置目标收藏等级。筛选结果符合条件，不代表那就是你一定想改的名单；预览还是值得认真看一遍。

想法终于可以往前走一点

早些时候，很多想法只能先放着。技术上看不懂的地方不少，也不知道该从哪里下手。

后来 Agent 越来越能帮上忙，原来搁着的东西开始有机会试一试。感觉：很快啊~

不过，做得更快不代表游戏里的每一种情况都已经确认。能加载、参数改了、模拟检查通过，和真实游戏里保存再读档后的结果，是不同的事情。还没核实的部分，我会尽量留在说明里，不因为有了一个新版本号就把它们抹掉。

这些模组还会继续慢慢整理。哪一处用起来顺手，哪一处还有问题，都可以在对应项目下面说。最好带上版本和发生时的步骤，这样更容易找到同一个问题。

Making mods started with playing games.

Small ideas kept coming up: a little more room around the base, more space in a container, fewer clicks to sort the Pals in storage. None of these had to be a huge problem. Running into them repeatedly was enough to make me want to try a change.

One change at a time

Some of the mods adjust limits: base radius, container capacity or the disassembler’s waiting queue. Others change a sequence of actions: sending a batch, finding materials across pages, or filtering Pals before setting a favorite level.

They may all reduce repetitive work, but they do different things. A larger disassembler queue does not make processing faster. A larger capacity for newly created boxes does not migrate every existing box in a save. Explaining that matters as much as making the change itself.

I prefer to keep these changes separate. You can choose the one you want without changing several other rules at the same time.

Keep the final decision

Batch actions should reduce clicks while leaving the actual choice visible. The condenser helper finds materials but does not press the final condensation button. The disassembler helper shows a list and waits for confirmation before sending it. Their consequences differ, so their prompts should too.

The favorite tool follows the same pattern: choose conditions, preview the matches, then set the target level. A list that matches a filter is not automatically the list you intended to change.

Getting ideas off the shelf

At first, plenty of ideas stayed unfinished because I didn’t understand how to approach them. As AI agents became more useful, I could try more of those ideas. “很快啊~” — that was quick~

Working faster doesn’t mean every in-game situation has been verified. Loading successfully, changing a parameter and passing isolated tests are different from saving and reloading in the actual game. I want the documentation to keep those distinctions visible.

If something feels useful, or something goes wrong, leave a note on the relevant project page. The version and steps you took will help make the report easier to follow.