D

Skill 详情

Dcc Cli Gateway

使用 dcc-mcp-cli 命令行控制实时 DCC 主机(包括 Maya、Blender、Houdini、Photoshop、3ds Max 及自定义工作室工具),兼容 ClawHub、OpenClaw、Cur...

来源平台:SkillHub
来源标识:SkillHub/dcc-cli-gateway
源文件:原始说明
AI 平台与模型 高关注 SkillHub 高 风险 下载 339 SkillHub
来源平台SkillHub
文档版本0.17.36
热度高关注
排名信号下载 339
概述 安装 文档 下载

快速判断

使用 dcc-mcp-cli 命令行控制实时 DCC 主机(包括 Maya、Blender、Houdini、Photoshop、3ds Max 及自定义工作室工具),兼容 ClawHub、OpenClaw、Cur...

最后校验2026-05-27
来源平台SkillHub
安全提示
下载副本ZIP 可用

适合任务

  • 按 SkillHub 收录说明复用成熟任务流程。
  • 通过下载包离线阅读完整 Skill 内容。
  • 结合热度指标优先评估常用 Skill。

输入与输出

输入:任务目标、上下文材料、文件路径、约束条件或需要处理的内容。

输出:按 Skill 说明生成的文档、代码、检查结果、计划、建议或操作步骤。

示例任务

  • 使用 Dcc Cli Gateway 帮我处理当前任务,并说明需要准备哪些输入。
  • 根据 Dcc Cli Gateway 的说明,先列出使用前的安全检查项。

安装方式

  1. 下载本站提供的 Skill ZIP 并解压。
  2. 把解压后的 Skill 目录放入当前 AI 工具支持的 skills 目录。
  3. 如需在线查看原始内容,可打开 GitHub 的 SKILL.md

在线原始地址:skillhub-dcc-cli-gateway/SKILL.md

风险边界

SkillHub 提供了源站安全报告入口,但本站不替代人工审查。使用前仍需检查权限、外部依赖和敏感数据边界。

SKILL.md 文档介绍

DCC CLI Gateway — Agent Control Plane

Use this skill when an agent host can run shell commands and should connect to

DCC-MCP through dcc-mcp-cli instead of MCP JSON-RPC. The CLI wraps the

gateway REST API and returns JSON by default. The bundled Python fallback sends

Accept: application/json because the gateway REST API itself now defaults to

compact TOON for agent-facing routes.

Connection order:

1. Use dcc-mcp-cli when it is already on PATH.

2. If missing, ask user permission, then download dcc-mcp-cli from GitHub Releases.

3. If the download fails, use the bundled Python stdlib REST fallback.

Install via OpenClaw/ClawHub, or point your agent at this SKILL.md after cloning

dcc-mcp-core/skills/dcc-cli-gateway/.

---

Critical Rules

| Situation | You MUST |

|-----------|----------|

| Starting any DCC task | Run python scripts/dcc_gateway.py health and python scripts/dcc_gateway.py list first |

| dcc-mcp-cli missing | Ask permission before --ensure-cli; fallback Python REST is allowed if download fails |

| Inventory returns total == 0 | Stop; do not run search, describe, or call |

| Gateway unreachable | Stop; explain; ask user permission before troubleshooting |

| User has not agreed to setup | Do not install packages, edit env files, launch GUI apps, or write configs |

| User approved setup | Follow [references/ZERO_INSTANCES_CLI.md](references/ZERO_INSTANCES_CLI.md) |

| After DCC crash/restart | Re-run list and search; old slugs may be invalid |

---

Configuration

dcc-mcp-cli and the Python helper read the gateway URL from DCC_MCP_BASE_URL.

export DCC_MCP_BASE_URL="${DCC_MCP_BASE_URL:-http://127.0.0.1:9765}"
dcc-mcp-cli health
python scripts/dcc_gateway.py health

For a one-off command:

python scripts/dcc_gateway.py --base-url http://127.0.0.1:9765 health

Quick probe helper:

python3 scripts/check_cli.py
py -3 scripts\check_cli.py

Flags: --base-url URL, --cli dcc-mcp-cli, --ensure-cli, --install-dir DIR, --pretty.

When the user approves downloading the CLI:

# Linux / macOS
python3 scripts/dcc_gateway.py --ensure-cli list
vx python scripts/dcc_gateway.py --ensure-cli list

# Windows
py -3 scripts\dcc_gateway.py --ensure-cli list
vx python scripts\dcc_gateway.py --ensure-cli list

Release assets are selected by platform:

| Platform | Asset |

|----------|-------|

| Windows x86_64 | dcc-mcp-cli-windows-x86_64.exe |

| Linux x86_64 | dcc-mcp-cli-linux-x86_64 |

| macOS Intel/Apple Silicon | dcc-mcp-cli-macos-universal2 |

If Python is not easy to locate, install vx first and run the helper through

vx python:

# Linux / macOS
curl -fsSL https://raw.githubusercontent.com/loonghao/vx/main/install.sh | bash

# Windows PowerShell
powershell -c "irm https://raw.githubusercontent.com/loonghao/vx/main/install.ps1 | iex"

---

Step 0 — Mandatory Instance Inventory

Run this every time you begin work or after the user starts/stops a DCC host:

python scripts/dcc_gateway.py health
python scripts/dcc_gateway.py list

Interpret dcc-mcp-cli list:

{
  "total": 1,
  "instances": [
    {
      "instance_id": "full-uuid",
      "instance_short": "a1b2c3d4",
      "dcc_type": "maya",
      "status": "available",
      "stale": false,
      "mcp_url": "http://127.0.0.1:8765/mcp"
    }
  ]
}

Report to the user:

1. total

2. Count by dcc_type

3. Any stale: true rows

4. The target instance_id or instance_short you will use

If total == 0, stop and ask whether the user wants setup guidance for the

target DCC. Continue only after explicit approval.

---

Step 1 — Search Tools

Only run this when inventory shows at least one non-stale target:

python scripts/dcc_gateway.py search --query sphere --dcc-type maya --limit 20

Copy the returned slug exactly. Gateway slugs look like:

maya.a1b2c3d4.maya_primitives__create_sphere

Never hand-build slugs.

---

Step 2 — Describe Schema

python scripts/dcc_gateway.py describe maya.a1b2c3d4.maya_primitives__create_sphere

Read tool.inputSchema and safety annotations before calling.

---

Step 3 — Call a Tool

python scripts/dcc_gateway.py call maya.a1b2c3d4.maya_primitives__create_sphere \
  --json '{"radius":2.0}'

Tool-specific fields (code, file_path, radius, and similar) belong inside

the --json object. Do not pass them as top-level CLI flags unless the CLI adds

an explicit first-class flag later.

See [references/CLI_CHEATSHEET.md](references/CLI_CHEATSHEET.md) for command

patterns and common errors.

---

What This Skill Does Not Use

  • MCP tools/list, tools/call, or resources/read
  • Raw curl workflows except when debugging the gateway itself
  • Direct Maya/Blender/Houdini scripting

The CLI is the preferred agent-facing control plane. The Python fallback uses

the same gateway REST endpoints only when the CLI is unavailable after a

download attempt fails.

建议反馈