Datapack authoring

You do not write Java. You write JSON.

A target file says “this structure can offer work” and points at four polls: rank, quest, reward, revenge. Polls are weighted lists. Quest templates are the actual jobs.

data/<your_namespace>/adventure_engine/
  targets/          polls/rank/
  ranks/            polls/quest/
  quest_templates/  polls/reward/
  polls/revenge/

That folder can live in a normal datapack, in kubejs/data/..., or under config/adventure_engine/. The config copy wins if the same file exists in the jar. /adventure reload rereads the overlay.

Easiest start: copy a target from the jar, change the structure list, keep adventure_engine:standard_rewards until you care about loot. Hook any structure id or tag, including other mods.

On quest poll entries, role is primary, side, or any. rank_min / rank_max use Scout=1 through Legend=5. Rank polls can also require distance from the box center, so a Champion job does not roll on a hut you are standing in the doorway of.

item_bands on find and sacrifice use those same 1–5 keys. The game rolls one item from the list. Find plants a marked stack in a chest (or drops it glowing). Sacrifice does not plant anything. The player has to bring the item.

Vanilla XP is minecraft_xp on the rank file. Leave the reward poll on standard_rewards until you have a reason not to.

Example of a pack hook: you have a custom dungeon mypack:quartz_keep. Copy the outpost target, point structures at that id, write one kill-leader template, and ship it. Players walk in, get a plaque, and the keep is now a job instead of an empty build.