Skip to content
Back to skills

Plugin Architect

ASecurity

构建 AI Skills/Plugins 的标准方法论 (Based on Anthropic Standard)

  • 321 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added May 26, 2026
toolspythongonextjsdatabase

Works with

  • mcp

Security analysis

A100/100

Scanned September 22, 2026

npx -y skills add kennyzir/7deer_skills --skill plugin-architect --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Plugin Architect?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Plugin Architect
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/kennyzir-plugin-architect/badge)](https://www.skillsdirectory.com/skills/kennyzir-plugin-architect)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: plugin-architect
description: 构建 AI Skills/Plugins 的标准方法论 (Based on Anthropic Standard)
metadata:
  version: "1.0"
---

# 🏗️ Plugin Architect Protocol

本 Skill 定义了本项目构建 **AI Plugins** 的唯一标准。所有 Skill 的创建与修改必须遵循此方法论。

## When This Skill Applies
*   当用户要求"创建一个新 Skill"时。
*   当需要重构现有 Skills 以符合标准时。
*   当需要为 Skill 添加外部工具 (MCP) 或快捷指令 (Slash Commands) 时。

## Instructions

一个完整的 **Plugin (Skill System)** 由四个核心组件构成。

### 1. The Identity (`plugin.json`)
插件的身份证。定义插件在系统中的唯一标识。
*   **Location**: `.agent/.claude-plugin/plugin.json` (Global) 或 `skill-name/plugin.json` (Local)
*   **Schema**:
    ```json
    {
      "schema_version": "1.0",
      "name": "devil-hunter-nexus-plugin",
      "description": "The cognitive engine and SEO factory for Roblox game sites.",
      "version": "1.0.0",
      "authors": ["Devil Hunter Team"]
    }
    ```

### 2. The Logic (`SKILL.md`)
AI 的思维链。必须包含明确的触发条件 (Trigger) 和结构化指令。

*   **Location**: `.agent/skills/<skill-name>/SKILL.md`
*   **Frontmatter**: Must include `name` and `description`.
*   **Mandatory Sections**:
    1.  `## When This Skill Applies`: 明确的触发条件 (Contextual Triggers)。
        *   *Good*: "When the user asks to analyze SEO metadata."
        *   *Bad*: "Use this for SEO."
    2.  `## Instructions`: 具体的操作步骤、原则或代码规范。
    3.  `## Actions` (Optional): 可用的工具或脚本引用。

### 3. The Interface (`commands/`)
用户的快捷入口。定义以 `/` 开头的显式调用指令。

*   **Location**: `.agent/commands/<command-name>.md`
*   **Format**:
    ```markdown
    ---
    description: Audit the current page for SEO compliance
    argument-hint: [url or path]
    active-skills: [nextjs-seo-foundations]
    ---
    
    # /audit-seo
    
    1. Check presence of Metadata (Title, Description).
    2. Validate JSON-LD Schema.
    3. Report missing 'canonical' tags.
    ```

### 4. The Tooling (`.mcp.json`)
外部能力的延伸。通过 Model Context Protocol 连接真实世界。

*   **Location**: `.agent/.mcp.json`
*   **Usage**: 如果 Skill 需要执行复杂脚本 (e.g., Python Script, Database Query),需在此注册 MCP Server。

---

## Workflow: How to Build a New Skill

1.  **Define Intent**: 是"思维模式" (Thinking) 还是"具体操作" (Action)?
    *   *Thinking* -> 重点编写 `SKILL.md` 的 Instructions。
    *   *Action* -> 重点编写 `commands/` 和 `scripts/`。

2.  **Create Directory**: `mkdir .agent/skills/<skill-name>`

3.  **Draft SKILL.md**:
    *   Copy metadata template.
    *   Write strict "When This Skill Applies" triggers.
    *   Write verifiable Instructions.

4.  **Register Command (Optional)**:
    *   If user needs frequent access, create `.agent/commands/<skill-shortcut>.md`.

5.  **Validation**:
    *   Ask AI: "Does this skill structure match the `plugin-architect` standard?"

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.