Minecraft Modpack Creator Skills

English | 简体中文

An Agent Skills workflow for building Minecraft modpacks with durable context, source-backed decisions, and repeatable integration reviews.

This repository is not a launcher, downloader, or mod manager. It is a skill set for AI agents that helps turn modpack development from ad hoc chat into a maintainable project workflow.

Use it when you want an agent to:

• Plan a new NeoForge / Minecraft modpack from zero.
• Add project memory to an existing modpack.
• Evaluate whether a candidate mod actually fits the current pack.
• Keep accepted decisions, rejected ideas, compatibility reports, and integration notes in files.

Why It Exists

The hard part of modpack development is rarely "find a mod." The hard part is keeping the reasoning clear:

• What experience is this pack trying to create?
• Which Minecraft version, loader, and server/client constraints matter?
• Which mods are accepted, rejected, or still pending?
• Does a new mod overlap with worldgen, progression, permissions, commands, config, or balance?
• Is the download source official, and where is the evidence?
• Where should the next AI session continue from?

These skills write that context into pack-state, so future work can continue from project state instead of a temporary conversation.

Skills

Each skill is a self-contained folder with a SKILL.md file.

Skill	Purpose
minecraft-pack-builder	Orchestrates the full workflow: requirement intake, candidate discovery, shortlist, analysis, compatibility review, source resolution, and state updates.
mod-analyzer	Analyzes one mod in depth: source validity, aliases, version / loader support, dependencies, configuration surface, implementation signals, and risks.
pack-compat	Decides whether one or more analyzed mods fit the current pack and records blockers, overlaps, and adoption conditions.
pack-context	Maintains the current pack context: goals, constraints, accepted / rejected mods, prior analyses, compatibility reports, and known risks.

Shared scripts and templates are owned by the minecraft-pack-builder skill. Other skills refer to those resources by skill and resource name, not by hard-coded repository paths.

Project State

The skills create and maintain a pack-state directory inside the modpack workspace:

pack-state/
  project-context.json
  mod-analyses/
  compat-reports/
  integration-docs/

When pack-state is missing, pack-context bootstraps it automatically. A minimal project-context.json starts like this:

{
  "packName": "my-pack",
  "minecraftVersion": null,
  "loader": "NeoForge",
  "distribution": "server-first",
  "goals": [],
  "hardConstraints": [],
  "acceptedMods": [],
  "rejectedMods": [],
  "pendingMods": [],
  "openCapabilityGaps": [],
  "knownRisks": [],
  "artifacts": {
    "modAnalyses": [],
    "compatReports": [],
    "integrationDocs": []
  }
}

Commit the generated pack-state directory to Git. It is the long-term memory of the pack.

Start a New Modpack

Create a clean workspace:

my-modpack/
  mods/
  config/

Then make this repository's skills/ available to your agent. Depending on your environment, that can mean installing the folders into the agent's skills directory or keeping this repository open as a workspace that supports local skills.

Give the agent a clear target. It should create pack-state itself if the directory is missing:

Use the minecraft-pack-builder skill.

I want to build a NeoForge server-first survival modpack from zero.
The direction is exploration, roads, towns, light automation, and long-term progression.
Avoid heavy client-only requirements and avoid mods that make worldgen significantly unstable.
Help me clarify the requirements, propose the first candidate mods, and update pack-state once decisions become stable.

Recommended loop:

1. Describe the gameplay goal and hard constraints.
2. Let pack-context initialize pack-state.
3. Let minecraft-pack-builder propose candidates.
4. Choose one or more candidate mods.
5. Let mod-analyzer write per-mod analysis artifacts.
6. Let pack-compat classify fit, blockers, and conditions.
7. Let pack-context update durable project state.

A good session ends with updated files, not just advice.

Connect an Existing Modpack

You do not need to rebuild an existing pack around these skills. Add a state layer and let the agent learn the current project shape.

First, make skills/ available to the agent. Then ask it to bootstrap pack-state from the existing files:

Use the pack-context skill.

This is an existing Minecraft modpack. Inspect the current workspace and create pack-state/project-context.json.
Use mods/, config/, defaultconfigs/, kubejs/, datapacks/, and any existing notes if present.
Separate confirmed facts from guesses. Do not invent version targets or design goals.

Backfill analysis for important mods gradually:

Use the mod-analyzer skill.

Analyze the most important worldgen, progression, and server gameplay mods in this pack.
Create one mod-analyses artifact per important mod, then use pack-compat to identify overlaps and risks.

For future changes, make the agent read pack-state first:

Use the minecraft-pack-builder skill.

I want to add a road or transportation system to this existing pack.
Read pack-state first, reuse prior analyses, propose candidate mods, and only recommend a mod when the official source and pack fit are clear.

Workflow Contract

• project-context.json is the current entry point for pack truth.
• mod-analyses/*.json records individual mod sources, capabilities, and risks.
• compat-reports/*.json records fit decisions, blockers, and adoption conditions.
• integration-docs/*.md records configuration, deployment, and maintenance notes for accepted mods.

If a decision will matter later, keep it out of chat-only memory.

What This Is Not

• Not a mod download site.
• Not a launcher.
• Not a dependency solver.
• Not a black box that decides the pack direction for you.

Its job is to help you make better decisions and preserve those decisions as project assets.

Repository Layout

skills/
  minecraft-pack-builder/
    SKILL.md
    scripts/
    references/templates/
  mod-analyzer/
    SKILL.md
  pack-compat/
    SKILL.md
  pack-context/
    SKILL.md

Principles

• Prefer official or trusted sources.
• Prefer explicit tradeoffs over vague "should be fine" conclusions.
• Keep accepted decisions, candidate ideas, and hypotheses separate.
• Reuse existing analysis instead of starting from scratch.
• Leave the project easier to continue than you found it.