Token导航 LogoToken导航

Agent Skill · 研究检索

minecraft-modding-workbench(我的世界模组工作台)

minecraft-modding-workbench 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

minecraft-modding-workbench 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为调用外部服务;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 65,612。

总安装 279周安装 12GitHub Stars 公开资料未说明下载 98核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:minecraft-modding-workbench(我的世界模组工作台)
来源仓库:https://github.com/adhi-jp/agent-skills
仓库路径:skills/minecraft-modding-workbench
安装命令:
npx skills add https://github.com/adhi-jp/agent-skills --skill minecraft-modding-workbench
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/adhi-jp/agent-skills --skill minecraft-modding-workbench

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

用于 Minecraft 模组开发的技术资源查找和工具推荐。可筛选 IDE 配置、调试工具和发布平台信息。需根据开发语言和目标平台选择合适工具链。建议优先使用社区验证过的成熟方案。minecraft-modding-workbench 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/minecraft-modding-workbench

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/adhi-jp/agent-skills --skill minecraft-modding…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要调用外部服务,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

minecraft-modding-workbench 怎么安装?

在终端执行 npx skills add https://github.com/adhi-jp/agent-skills --skill minecraft-modding-workbench。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

minecraft-modding-workbench 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

minecraft-modding-workbench 需要什么权限?

权限级别是调用外部服务。该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

minecraft-modding-workbench 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

minecraft-modding-workbench 用的人多吗?

公开数据是累计安装 279 次,近一周安装 12 次,下载 98 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

minecraft-modding-workbench 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
minecraft-modding-workbench

我的世界模组工作台 · 本页

279—调用外部服务未明确有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Minecraft Modding Workbench

Support fast, version-aware Minecraft modding. Treat the minecraft-modding MCP server as the primary source of truth, then turn verified findings into working code and assets.

Scope

  • Supports Fabric, NeoForge, and Architectury.
  • Requires the minecraft-modding MCP server from @adhisang/minecraft-modding-mcp.
  • Prefer project-aware MCP calls when a workspace exists. Reuse the repository root as projectPath.
  • Use the high-level v3 entry tools first: inspect-minecraft, analyze-symbol, compare-minecraft, validate-project, analyze-mod, and manage-cache.
  • Covers Forge-style access transformers through validate-project (task access-transformer) for NeoForge, in addition to Fabric-style access wideners.
  • Use the NBT helpers (nbt-to-json, json-to-nbt, nbt-apply-json-patch) when working with level.dat, chunk, playerdata, or command-driven NBT. Stay in typed JSON while editing and re-encode once at the end.

Default Behavior

  • Produce runnable feature slices, not generic advice.
  • Infer loader, version, mappings, modid, package, Java version, and project conventions from the workspace before asking questions.
  • Ask only the minimum blocking question when the workspace is absent or contradictory.
  • Prefer explicit TODOs or placeholder assets over stalling on art or balance details.
  • When the user clearly requests implementation rather than explanation, default to delivering code.
  • Respond in the user's language when practical, but keep the workflow and trigger logic language-agnostic.

First Pass

  1. Detect the project shape.

- Read gradle.properties, build.gradle, build.gradle.kts, settings.gradle, fabric.mod.json, neoforge.mods.toml, mixin configs, and nearby registration classes. - Infer loader, Minecraft version, mappings, Java version, modid, base package, whether the project already uses datagen, and whether the workspace is single-loader or Architectury multi-module. - Record the workspace root you will pass as projectPath to MCP tools.

  1. Read the existing code before writing new code.

- Match the project's naming, package layout, registration helpers, and client/server split. - Reuse existing registries, tabs, packet patterns, and datagen providers when present. - Prefer workspace-aware MCP resolution before manual version or mapping selection.

  1. Load only the relevant references.

- Read references/fabric.md for Fabric projects. - Read references/neoforge.md for NeoForge projects. - Read references/architectury.md for Architectury multi-module projects. - Read references/bootstrap-from-template.md when the workspace is a sparse template. - Read references/task-checklists.md for the requested feature slice. - Read references/mcp-recipes.md when payload shape, recovery, or sequencing is unclear.

  1. Find the closest vanilla example before implementing behavior.

- Start with inspect-minecraft or analyze-symbol. - Drop to low-level tools only when the high-level answer still leaves the implementation ambiguous.

If no project exists yet, ask only for loader, Minecraft version, modid, and package name. If the task depends on an external generator or template that is not present, say so explicitly instead of fabricating generated files.

MCP Guardrails

  • Start with the highest-level read-only MCP call that can answer the question.

- inspect-minecraft: versions, artifacts, vanilla classes, source search, raw files. - analyze-symbol: existence, mappings, lifecycle, workspace compile-time names, API overview. - compare-minecraft: migration and registry/class diffs. - validate-project: workspace, Mixin, access widener, and Forge-style access transformer validation. - analyze-mod: mod JAR summary, search, decompile, remap preview/apply. - manage-cache: stale cache or index diagnosis, including the verify action and preview-then-apply maintenance.

  • Reach for these supporting utilities directly when the entry tools do not cover the job:

- get-registry-data: structured registry bodies (blocks, items, biomes, …) via the server data generator for one version. - get-runtime-metrics: service counters and latency snapshots when cache, search, or index behaviour looks off. - nbt-to-json, json-to-nbt, nbt-apply-json-patch: typed-JSON round-trip and RFC6902-style in-place edits for Java Edition NBT payloads.

  • Drop to low-level tools only for exact code, exact descriptors, raw registry bodies, detailed validator output, or direct JAR/remap control.
  • Keep version and mapping discipline.

- Pass projectPath, preferProjectVersion=true, and preferProjectMapping=true when supported. - Still pass explicit version to tools that require it, such as validate-mixin, validate-access-widener, and resolve-workspace-symbol. - Treat artifact-backed lookup as a two-step flow: resolve-artifact first, then find-class, search-class-source, get-artifact-file, or list-artifact-files.

  • Parallelize only independent read-only discovery calls once projectPath, loader, version, and mapping are known.

- Keep dependent chains sequential. - Do not run manage-cache, index-artifact, or remap/mutating flows in parallel with calls that depend on the same cache or JAR.

  • If payload shape is unclear or an entry tool errors, read references/mcp-recipes.md before inventing fields or dropping to a lower-level tool.

Unsupported or Risky Requests

  • Do not silently treat Quilt or legacy Forge as Fabric, NeoForge, or Architectury.
  • For legacy Forge-only or other unsupported loaders, limit help to verified workspace facts, logs, and migration boundaries. Say that full guidance is outside this skill.
  • If MCP is unavailable, misconfigured, or stale, say so immediately, fall back to workspace and log inspection, and keep any fix narrow. The same rule covers version skew: if a tool, task, or argument this skill names (for example, manage-cache action: "verify", validate-project task access-transformer, analyze-symbol api-overview / exact-map, the nested inspect-minecraft subject.kind: "artifact" shape, or the NBT helpers) is rejected as unknown, treat it as evidence that the installed MCP is older than what this skill's recipes target, say so explicitly, and route the request through the nearest older-compatible tool or a workspace-only fallback rather than fabricating a different payload shape.
  • If workspace files contradict the prompt, call out the contradiction and resolve it from checked files before coding.
  • If the request depends on a symbol, event, registry entry, or vanilla hook you cannot verify, say that it is unverified or unsupported instead of inventing it. Offer the closest verified alternative.

Core Workflow

  1. Inspect vanilla or existing mod code that already solves the same problem.
  2. Translate that pattern into the user's loader, module boundary, and mapping namespace.
  3. If the template is too empty, bootstrap the missing project skeleton first.

- Add only the minimum entrypoints, registration classes, client hooks, and datagen wiring needed for the requested feature. - Do not create every possible system up front.

  1. Implement the whole slice in one pass.

- Include registrations. - Include client wiring when needed. - Include required JSON resources or datagen hooks. - Include lang keys, loot tables, blockstates, models, tags, recipes, or screen wiring when the feature needs them. - In Architectury workspaces, keep shared gameplay logic in common and loader-specific wiring in platform modules unless the workspace already uses another verified pattern.

  1. Run the verification loop before calling the task done.
  2. Report assumptions, placeholders, and follow-up tasks briefly.

Delivery Rules

  • Match the current project style before introducing a new abstraction.
  • Do not invent mapping names, event names, registration order, or descriptors. Verify them.
  • Prefer stable loader APIs or events over Mixins when the loader already exposes a clean hook.
  • When the project is template-only, create the smallest working scaffold that can compile and host the requested feature.
  • In Architectury projects, keep code in common by default and move only loader-bound code to fabric or neoforge.
  • In Architectury templates that already route both loaders through a shared init method, do not add no-op platform edits just to mirror a shared content change.
  • Use @ExpectPlatform, Architectury abstractions, or a plain Java interface/service split only when the code truly needs platform-specific behavior.
  • Keep side separation correct. Put renderer, screen, and other client-only code behind the proper client entrypoint or event.
  • Prefer datagen when the request creates repeated JSON or more than a couple of content entries.
  • Preserve existing helper classes, registries, and package structure instead of replacing them wholesale.
  • Keep fixes narrow during debugging. Identify the concrete failure first, then patch the cause.

Verification Loop

Run verification as part of the default workflow, not as an optional extra.

  1. Run ./gradlew build after structural code changes.
  2. Run the loader's datagen task when datagen was added or updated, or when generated resources are the project's normal asset path.

- Fabric: ./gradlew runDatagen - NeoForge: run the project's configured datagen task or run configuration. - Architectury: run the root build and the relevant platform datagen task when the workspace defines one.

  1. Run a client launch when the change touches rendering, menus, screens, entity models, or runtime-only behavior and the environment allows it.

- Use the project's existing runClient task or equivalent. - In Architectury workspaces, prefer the platform-specific client run task that exercises the changed module.

  1. Run or extend automated tests when the project already has them or when the new logic is isolated enough to justify them.

- Prefer existing GameTests, loader test harnesses, or integration tests for gameplay behavior. - Add focused unit tests for pure Java helpers, codecs, serializers, or data transforms.

  1. If a command cannot run in the current environment, say so explicitly and still perform static validation with MCP tools and code inspection.

- In sandboxed environments, retry Gradle with a writable GRADLE_USER_HOME before treating home-directory lock or cache failures as project issues.

At minimum, aim to leave the project in a state that passes build or has a concrete, localized reason why it cannot.

Fast Debugging Order

  • Mixin crash: start with validate-project, then validate-mixin if you need exact issue detail. Confirm the owner, method name, descriptor, and mapping namespace before patching code.
  • Access widener failure (Fabric): start with validate-project, then validate-access-widener with an explicit version. Confirm that the header namespace matches the entry names.
  • Access transformer failure (NeoForge): start with validate-project (task access-transformer), then validate-access-transformer with an explicit version and atNamespace. Confirm the file's entry namespace matches what the workspace expects (usually mojang on modern NeoForge, srg on legacy projects).
  • Registry or missing-content issue: inspect the existing registration flow, confirm registry IDs, then check required resource files. get-registry-data returns the vanilla-version entry list only; absence from its output is not evidence of a missing modded, dependency, or datapack entry, so fall back to workspace registration code, dependency metadata, and datagen output for those cases.
  • NBT payload corruption or schema drift: decode with nbt-to-json, edit in typed JSON (or nbt-apply-json-patch), preserve DataVersion, then re-encode with json-to-nbt using matching compression.
  • Cache or index anomalies: read get-runtime-metrics before mutating anything, then run manage-cache with action: "verify" in preview mode before prune, rebuild, or delete.
  • Texture or model issue: verify resource location casing, JSON paths, generated assets, and item-block model linkage.
  • Side-only crash: inspect client init, renderer registration, and level.isClientSide() or equivalent boundaries.
  • Porting failure: start with compare-minecraft, then diff the affected class signatures, then update mappings and loader-specific APIs.

When one of these categories matches, read the corresponding section in references/task-checklists.md and the loader-specific Common Pitfalls section before broad rewrites.

References

  • Fabric patterns: references/fabric.md
  • NeoForge patterns: references/neoforge.md
  • Architectury patterns: references/architectury.md
  • Template bootstrap patterns: references/bootstrap-from-template.md
  • Delivery checklists by task shape: references/task-checklists.md
  • MCP payload and recovery recipes: references/mcp-recipes.md
  • For current upstream migration guidance, consult the official Fabric, NeoForge, and Architectury docs or release notes that match the target loader and Minecraft version instead of relying on hardcoded URLs.

平台分布

Codex

33.63%
按下载量换算33

Claude

31.79%
按下载量换算31

Cursor

18.08%
按下载量换算18

Gemini CLI

8.43%
按下载量换算8

安全审计

Gen Agent Trust Hub

通过

Socket

可疑

Snyk

可疑

权限和风险

调用外部服务

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills