Token导航 LogoToken导航TokenDH.com
研究检索执行命令github未标认证来源可访问许可证需确认审计通过

devstudiodevstudio 搜索

Agent Skill

devstudio 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。

总安装

343

周安装

14

GitHub Stars

11

下载量

110
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

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:devstudio(devstudio 搜索)
来源仓库:https://github.com/lobbi-docs/claude
仓库路径:skills/devstudio
安装命令:
npx skills add https://github.com/lobbi-docs/claude --skill devstudio
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/lobbi-docs/claude --skill devstudio

简介

为 Claude Code 插件开发提供集成测试与热重载支持环境。

  • 支持插件清单验证、依赖图可视化和回归测试套件构建。
  • 包含四个子系统:构建、测试、验证与文档生成模块。
  • 适用于插件迭代开发与发布前的质量保障流程。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。
  • devstudio 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Plugin Dev Studio Workflow

Comprehensive development environment for building, testing, and validating Claude Code plugins with hot-reload support.

When to Use This Skill

Activate this skill when:

  • Developing a new plugin and need rapid iteration with live validation
  • Debugging plugin manifest or resource file issues
  • Visualizing plugin dependency graphs for architecture review
  • Building regression test suites for plugin commands
  • Preparing a plugin for publication to a registry

Architecture Overview

The Dev Studio consists of four core subsystems:

+------------------+     +---------------+     +-------------------+
|   FileWatcher    | --> |  HotReloader  | --> | Resource Registry |
| (FNV-1a hashing) |     | (Validation)  |     | (Live state)      |
+------------------+     +---------------+     +-------------------+
                                |
                                v
                     +--------------------+
                     | Console Reporter   |
                     | (file:line errors) |
                     +--------------------+

+-------------------+     +---------------------------+
| PluginPlayground  |     | DependencyGraphRenderer   |
| (Record/Replay)   |     | (ASCII + Mermaid output)  |
+-------------------+     +---------------------------+

FileWatcher

Monitors a plugin directory for file changes using filesystem watchers with content-hash-based change detection.

Key design decisions:

  • Uses FNV-1a (32-bit) hashing for speed — non-cryptographic, but extremely fast for small files
  • Only triggers reload when file content actually changed (ignores timestamp-only updates)
  • Debounces rapid changes with a 100ms window to batch editor save operations
  • Classifies files by resource type (command, skill, agent, config, source)

HotReloader

Processes file changes and maintains a live resource registry.

On each change:

  1. If manifest changed: re-read, re-validate JSON structure and required fields
  2. If markdown resource changed: re-lint frontmatter (YAML validity, required fields)
  3. Update resource registry: add new resources, update modified ones, remove deleted ones
  4. Report validation results with file:line references for inline error display

PluginPlayground

Isolated execution context for testing plugin commands.

Record-replay workflow:

  1. Register mock capabilities for external dependencies
  2. Execute commands and record full input/output pairs as fixtures
  3. Save fixtures to tests/fixtures/ as JSON
  4. Replay fixtures in CI to detect regressions

DependencyGraphRenderer

Builds and renders plugin dependency graphs.

Two output formats:

  • ASCII tree: Uses Unicode box-drawing characters for terminal display
  • Mermaid diagram: Pasteable into GitHub markdown, Notion, or Mermaid Live Editor

Development Workflow

Phase 1: Scaffold and Configure

# Create plugin structure
mkdir -p my-plugin/{commands,skills,agents,config,src,tests/fixtures}
mkdir -p my-plugin/.claude-plugin

# Create minimal manifest
cat > my-plugin/.claude-plugin/plugin.json << 'EOF'
{
  "name": "my-plugin",
  "version": "0.1.0",
  "description": "My new plugin",
  "contextEntry": "CONTEXT.md",
  "capabilities": {
    "provides": ["my-capability"],
    "requires": []
  }
}
EOF
# Create required operator runbook
cat > my-plugin/CLAUDE.md << 'EOF'
# My Plugin Guide

## Purpose
- Brief description of plugin intent.

## Supported Commands
- command-name (commands/command-name.md)

## Prohibited Actions
- List destructive or out-of-scope operations.

## Required Validation Checks
- npm run check:plugin-context
- npm run check:plugin-schema

## Context Budget
1. CONTEXT_SUMMARY.md
2. commands/index (or commands/)
3. README.md and only task-relevant deep docs

## Escalation Path
- Describe who reviews risky or blocking changes.
EOF
# Create operator context entrypoint (keep concise)
cat > my-plugin/CONTEXT.md << 'EOF'
# my-plugin Context

## Purpose
One-paragraph operator summary.

## Key Commands
- /my:command

## Agent Inventory
- my-agent

## Load Deeper Docs When
- You need implementation details or architecture rationale.
EOF

Phase 2: Develop with Hot-Reload

# Start the dev server with file watching
/mp:dev serve ./my-plugin --watch

The dev server will:

  • Validate the manifest on startup
  • Discover and register all commands, skills, and agents
  • Watch for file changes and re-validate on save
  • Show inline errors with file:line references

Phase 3: Test in Playground

# Launch interactive playground
/mp:dev playground ./my-plugin

# In the playground:
# > run /my:command "test input"
# > save my-test-case
# > log

Phase 4: Visualize Dependencies

# Show dependency graph
/mp:dev graph ./my-plugin

Review the ASCII tree to verify:

  • All capability declarations are correct
  • Inter-resource dependencies are properly linked
  • No circular dependencies exist

Phase 5: Validate Before Publish

# Full validation suite
/mp:dev validate ./my-plugin

Fix all errors before publishing. Warnings are advisory but should be addressed.

Phase 6: Regression Testing

# List saved fixtures
/mp:dev fixture list

# Replay a specific fixture
/mp:dev fixture replay my-test-case

Validation Rules

Manifest Validation (plugin.json)

RuleSeverityDescription
Valid JSONerrorFile must be parseable JSON
name fielderrorMust be a non-empty string
version fielderrorMust be a non-empty string
description fielderrorMust be a non-empty string
contextEntry fielderrorMust reference CONTEXT.md or PLUGIN_CONTEXT.md
CLAUDE.md presenterrorPlugin root must include CLAUDE.md runbook
capabilities presentwarningPlugin should declare capabilities
capabilities.provides is arrayerrorMust be an array of strings
capabilities.requires is arrayerrorMust be an array of strings

Command/Skill Validation (.md files)

RuleSeverityDescription
Has frontmattererrorMust start with ---
Frontmatter closederrorMust have closing ---
name in frontmattererrorCommands and skills must have a name
description in frontmatterwarningShould have a description
Top-level headinginfoShould have # Title after frontmatter

File Structure

plugin-root/
  .claude-plugin/
    plugin.json          # Plugin manifest (validated by HotReloader)
  CLAUDE.md              # Required operator runbook and context budget
  CONTEXT.md             # Minimal operator context entrypoint
  commands/
    *.md                 # Slash commands (validated for frontmatter)
  skills/
    */SKILL.md           # Skills (validated for frontmatter)
  agents/
    *.md                 # Agents (validated for frontmatter)
  config/
    *.json               # Configuration files
  src/
    devstudio/
      types.ts           # Type definitions
      server.ts          # FileWatcher, HotReloader, Playground, GraphRenderer
  tests/
    fixtures/
      *.json             # Recorded test fixtures (from playground)

Troubleshooting

Common Issues

"Plugin manifest not found"

  • Ensure .claude-plugin/plugin.json exists in the plugin root
  • Check that the path passed to /mp:dev is correct

"Missing YAML frontmatter"

  • Markdown resource files must start with --- on the first line
  • Frontmatter must be closed with a second --- line

"Content hash unchanged but file was modified"

  • The FileWatcher uses content hashing, not timestamps
  • If only whitespace changed, verify the hash actually differs
  • The FNV-1a hash has a very low (but non-zero) collision rate for small changes

"Fixture replay fails with empty output"

  • Playground execution is a simulation; real command execution requires the Claude runtime
  • Record the actual output using recordOutput() after command execution

Source Files

  • Types: src/devstudio/types.ts
  • Implementation: src/devstudio/server.ts
  • Command: commands/dev.md
  • This skill: skills/devstudio/SKILL.md

Skill version: 1.0 Module: dev-studio (marketplace-pro plugin)

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

37.55%
按下载量换算41

Claude

26.33%
按下载量换算29

Cursor

17.05%
按下载量换算19

Gemini CLI

9.86%
按下载量换算11

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

执行命令

安装流程涉及命令执行,可能通过 npx skills add https://github.com/lobbi-docs/claude --skill devstudio 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

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。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills