Use this skill for biome, dimension, feature, or structure data and their\nregistration. Use minecraft-datapack for non-worldgen data and\nminecraft-modding for non-worldgen gameplay code.
Use when: the task changes worldgen data, registration, or injection.Do not use when: the task is non-worldgen datapack work (minecraft-datapack).Do not use when: the task is non-worldgen mod systems (minecraft-modding).| Approach | \nBest When | \nPlatform | \n
|---|---|---|
| Datapack JSON | \nChange data supplied by a pack | \nVanilla, any server | \n
| Mod + Datagen | \nRegistering new biomes/dimensions, code-driven | \nNeoForge / Fabric | \n
| Biome Modifier (NeoForge) | \nAdding features/spawns to existing biomes | \nNeoForge | \n
| BiomeModification API (Fabric) | \nAdding features/spawns to existing biomes | \nFabric | \n
Worldgen registries are datapack registries: their files load at world load and\ntheir registry path determines the data path. Read the NeoForge registry\nguide before choosing a\nmod-specific registry path.
\nTreat Minecraft 26.x as the current lane for new work. Use Java 25 and start\neach JSON schema from the exact target's vanilla data or generated output. Do\nnot copy a 1.21 shape into a 26.x pack merely because it parses as JSON.
\nPreserve an established 1.21.x project on Java 21 and its matching schema unless\nthe task explicitly includes an upgrade. Do not mix compatibility lanes.
\nThe 26.1 migration primer\nremoves minecraft:random_patch and minecraft:no_bonemeal_flower. It replaces\nthe random-patch pattern with a separate minecraft:simple_block configured\nfeature and placements for count, random offset, and block-predicate filtering.\nInspect the relevant primer section before migrating code or data.
Read legacy 1.21 JSON patterns only\nwhen the project targets that compatibility lane. Those examples are not\nrelease artifacts for 26.x.
\ndata/<namespace>/\n├── worldgen/\n│ ├── biome/\n│ │ └── my_biome.json\n│ ├── configured_feature/\n│ │ └── my_ore.json\n│ ├── placed_feature/\n│ │ └── my_ore_placed.json\n│ ├── noise_settings/\n│ │ └── my_dimension_noise.json\n│ ├── structure/\n│ │ └── my_structure.json\n│ ├── structure_set/\n│ │ └── my_structures.json\n│ ├── processor_list/\n│ │ └── my_processors.json\n│ ├── template_pool/\n│ │ └── my_pool.json\n│ └── carver/\n│ └── my_carver.json\n├── dimension/\n│ └── my_dimension.json\n├── dimension_type/\n│ └── my_type.json\n├── tags/\n│ └── worldgen/\n│ └── biome/\n│ └── is_forest.json\n└── neoforge/\n └── biome_modifier/ (NeoForge mod only)\n └── add_ores.json\nBuild and review the graph from its leaves upward:
\nUse fully qualified identifiers across namespaces. An external minecraft: or\ndependency reference is valid when that dependency supplies the registry entry;\ndo not create a local copy merely to satisfy static checking. If the same pack\ncontains that external namespace and registry directory, treat it as local and\nverify the target exists.
For 26.x biome and dimension data, use the exact target's vanilla data or\ndatagen output as the schema source. The older effects and dimension-type\nfields do not model newer environment behavior. The 11 decoration steps still\norganize placed features; choose the semantically appropriate step and keep ore\nplacement in underground_ores.
The version-labeled 1.21.5 biome and dimension examples are in\nlegacy 1.21 JSON patterns.
\nFor the 26.1 replacement for a simple random patch, the migration primer shows\na simple_block configured feature and a placed feature with count, random\noffset, and a block-predicate filter. Adapt the exact values and block state to\nthe target release's generated data.
At data/<namespace>/worldgen/configured_feature/my_plant.json:
{\n \"type\": \"minecraft:simple_block\",\n \"config\": {\n \"to_place\": {\n \"type\": \"minecraft:simple_state_provider\",\n \"state\": { \"Name\": \"minecraft:sweet_berry_bush\", \"Properties\": { \"age\": \"3\" } }\n }\n }\n}\nAt data/<namespace>/worldgen/placed_feature/my_plant.json:
{\n \"feature\": \"<namespace>:my_plant\",\n \"placement\": [\n { \"type\": \"minecraft:count\", \"count\": 96 },\n {\n \"type\": \"minecraft:random_offset\",\n \"xz_spread\": { \"type\": \"minecraft:trapezoid\", \"min\": -7, \"max\": 7, \"plateau\": 0 },\n \"y_spread\": { \"type\": \"minecraft:trapezoid\", \"min\": -3, \"max\": 3, \"plateau\": 0 }\n },\n {\n \"type\": \"minecraft:block_predicate_filter\",\n \"predicate\": {\n \"type\": \"minecraft:all_of\",\n \"predicates\": [\n { \"type\": \"minecraft:matching_block_tag\", \"tag\": \"minecraft:air\" },\n { \"type\": \"minecraft:matching_blocks\", \"blocks\": \"minecraft:grass_block\", \"offset\": [0, -1, 0] }\n ]\n }\n }\n ]\n}\nBiome modifiers load from\ndata/<modid>/neoforge/biome_modifier/<path>.json. They can target a biome id\nor tag and add or remove placed features, among other changes. The current\nBiome Modifiers guide\ndocuments their schemas, decoration steps, and datagen.
For neoforge:add_features, features accepts a placed-feature id, list, or\ntag. Vanilla placed features may be referenced in biome JSON or added with a\nmodifier, but NeoForge cautions against doing both because feature-order cycles\ncan crash world loading. Prefer a copy under the mod namespace when an injected\nvanilla feature would create that risk.
When targeting a biome from an optional dependency, put the target in a biome\ntag entry with required: false, then use that tag in the modifier. This lets\nthe pack load when the dependency is absent.
For any current release, derive structure, template-pool, dimension, and\ndimension-type JSON from that release's vanilla data or datagen output. Confirm\nthe reference graph before launching a test world:
\nstructure_set references structure.start_pool references template_pool; each single-pool element\nreferences its structure template and processor list.dimension.type references dimension_type; a noise generator's string\nsettings references worldgen/noise_settings.For Fabric registration or mod datagen, use the exact loader and API version's\ndocumentation rather than copying 1.21 code into a 26.x project.
\nThe detailed 1.21 structure and dimension examples are in\nlegacy 1.21 JSON patterns.
\ndata/<namespace>/worldgen/ (or equivalent mod resources path)../scripts/validate-worldgen-json.sh --root /path/to/datapack-or-mod-resources\n# Strict mode treats warnings as failures:\n./scripts/validate-worldgen-json.sh --root /path/to/datapack-or-mod-resources --strict\nworldgen/** and neoforge/biome_modifier/**placed_feature -> configured_featurestructure_set -> structure and biome/biome_modifier feature targetsjigsaw structure -> start_pool and template_pool -> structure template / processor_listcarvers,\neffects, or dimension settings match that release's schema./locate structure <namespace>:my_structure\n/locate biome <namespace>:my_biome\n/place feature <namespace>:my_ore\nplace feature takes a configured-feature ID, not its placed-feature wrapper.\nSee Mojang's place command reference in the 1.19 release notes./execute in (dimension must exist at world load, not added via /reload):\nexecute in <namespace>:my_dimension run tp @s 0 100 0\nlatest.log for worldgen errors (missing biome references, malformed noise settings)./reload refreshes datapack JSON but does not re-generate already-generated chunks. Test new worldgen in a fresh world or newly generated chunks. For existing test worlds, use a disposable copy and a purpose-built chunk reset/regeneration workflow; /fill only replaces blocks and is not a substitute for world generation.