Agent助手
- 作者仓库星标 0
- 作者更新于 实时读取
- 作者仓库 skills-registry
- 领域
- AI 智能
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 88 / 100 · 社区维护
- 作者 / 版本 / 许可
- @tomevault-io · 未声明 license
- Token 消耗评级
- 低消耗
- 接入复杂程度
- 需简单配置
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- macOS · Linux · Windows
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- Shell 执行
- 网络行为
- 仅限本地
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生,可能与作者实际意图存在差异。
需要注意: 未限定 allowed-tools,默认拥有全部工具权限。
---
name: agent-factory
description: Claude Code agent generation system that creates custom agents and sub-agents with enhanced YAML…
category: AI 智能
runtime: 无特殊运行时
---
# agent-factory 输出预览
## PART A: 任务判断
- 适用问题:提示词、Agent 工作流、模型评估或自动化推理。
- 输入要求:目标材料、限制条件、期望输出和验收方式。
- 证据边界:围绕“What This Skill Does / Capabilities / Agent Types Supported”读取原文规则,不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断:确认你的需求是否属于提示词、Agent 工作流、模型评估或自动化推理,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“What This Skill Does / Capabilities / Agent Types Supported”拆成步骤,说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果:给出可复制的命令、文件改动、检查清单或内容草稿,并说明如何继续迭代。
- **04** 风险边界:结合 读取文件、写入/修改文件、执行终端命令、主要在本地完成、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、写入/修改文件、执行终端命令;主要在本地完成;通常不需要额外 API Key。
- 先小样例验证,再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 原文没有稳定的斜杠命令要求。安装验证后通常全局生效,直接在对话里点名这个 Skill 并描述任务即可。
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是:读取文件、写入/修改文件、执行终端命令。
先用一个小任务确认它会围绕“What This Skill Does / Capabilities / Agent Types Supported”工作;涉及文件或命令时,先看 diff、日志、预览或测试结果。
检查最终产物是否包含明确结果、必要证据和下一步动作;如果输出泛泛而谈,就补充输入、边界和验收标准后重跑。
---
name: agent-factory
description: Claude Code agent generation system that creates custom agents and sub-agents with enhanced YAML…
category: AI 智能
source: tomevault-io/skills-registry
---
# agent-factory
## 什么时候使用
- 把 AI / Agent方向的常用动作沉淀成 Agent 可调用的技能 适合处理AI Agent、提示词、模型评估与自动化推理,核心价值是把输入、判断、执行、验证和交付边界固定下来,避免 Agent 泛泛回答。 把任务拆成可执行、可检查…
- 面向提示词、Agent 工作流、模型评估或自动化推理,优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「What This Skill Does / Capabilities / Agent Types Supported」组织步骤,不把推断写成作者事实。
- 读取文件、写入/修改文件、执行终端命令;主要在本地完成;通常不需要额外 API Key。
- 先跑小样例,确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown,不编造命令、平台或依赖。 作者原文负责流程事实;仓库文件负责来源和命令;流狐只补充适用场景、限制和质量判断。
skill "agent-factory" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> What This Skill Does / Capabilities / Agent Types Supported
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、写入/修改文件、执行终端命令 | 主要在本地完成
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
} Agent Factory
A comprehensive system for generating production-ready Claude Code agents and sub-agents. This skill provides templates, standards, and generation tools to create custom agents that seamlessly integrate with Claude Code's agent system.
What This Skill Does
This skill helps you create custom Claude Code agents for any domain or workflow. It generates properly formatted agent files that Claude Code can automatically discover and invoke when relevant.
Capabilities
- Generate Custom Agents - Create specialized agents for any domain (frontend, backend, testing, product, etc.)
- Enhanced YAML Frontmatter - Rich metadata including color coding, field categorization, expertise levels
- Tool Access Guidance - Recommends optimal tool configurations based on agent type
- MCP Integration - Suggests relevant MCP server tools for enhanced capabilities
- Execution Pattern Assignment - Ensures proper parallel/sequential execution for safety
- Validation - Checks agent configuration against best practices
Agent Types Supported
Strategic Agents (Lightweight, Parallel-Safe)
- Purpose: Planning, research, analysis
- Tools: Read, Write, Grep only
- Execution: 4-5 agents can run in parallel
- Color: Blue
- Examples: product-planner, market-researcher, architect
Implementation Agents (Full Tools, Coordinated)
- Purpose: Code writing, feature building
- Tools: Read, Write, Edit, Bash, Grep, Glob
- Execution: 2-3 agents coordinated
- Color: Green
- Examples: frontend-developer, backend-developer, api-builder
Quality Agents (Heavy Bash, Sequential Only)
- Purpose: Testing, validation, review
- Tools: Read, Write, Edit, Bash, Grep, Glob
- Execution: 1 agent at a time (NEVER parallel)
- Color: Red
- Examples: test-runner, code-reviewer, security-auditor
Coordination Agents (Lightweight, Orchestration)
- Purpose: Manages other agents, validates integration
- Tools: Read, Write, Grep
- Execution: Orchestrates others
- Color: Purple
- Examples: fullstack-coordinator, workflow-manager
Enhanced YAML Frontmatter
Every generated agent includes rich metadata:
---
name: agent-name-kebab-case
description: When to invoke this agent
tools: Read, Write, Edit # Comma-separated
model: sonnet # sonnet|opus|haiku|inherit
color: green # Visual categorization
field: frontend # Domain area
expertise: expert # beginner|intermediate|expert
mcp_tools: mcp__playwright # MCP integrations
---
Field Categories
Development: frontend, backend, fullstack, mobile, devops
Quality: testing, security, performance
Strategic: product, architecture, research, design
Domain: data, ai, content, finance, infrastructure
Color Coding
- Blue: Strategic/planning agents
- Green: Implementation/development agents
- Red: Quality/testing agents
- Purple: Coordination/orchestration agents
- Orange: Domain-specific specialists
Expertise Levels
- Beginner: Simple, focused tasks
- Intermediate: Moderate complexity workflows
- Expert: Advanced, complex operations
How to Use
Quick Start
- Open the prompt template: documentation/templates/AGENTS_FACTORY_PROMPT.md
- Scroll to bottom - Find template variables
- Fill in your details:
AGENT_NAME: my-custom-agent DESCRIPTION: What this agent does and when to invoke it DOMAIN_FIELD: frontend TOOLS_NEEDED: Read, Write, Edit, Bash - Copy entire prompt - Include filled variables
- Paste into Claude - Claude.ai, Claude Code, or API
- Receive agent file - Complete .md file ready to use
- Install agent - Copy to
.claude/agents/or~/.claude/agents/
Example Invocation
@agent-factory
Create a custom agent:
Name: api-integration-specialist
Type: Implementation
Domain: backend
Description: API integration expert for third-party services
Capabilities: OAuth, REST clients, error handling
Tools: Read, Write, Edit, Bash
MCP: mcp__github
Output: Complete .claude/agents/api-integration-specialist.md file
Generated Agent Structure
Each generated agent is a single Markdown file:
---
name: custom-agent
description: Triggers auto-invocation
tools: Read, Write, Edit
model: sonnet
color: green
field: backend
expertise: expert
mcp_tools: mcp__github
---
You are a [role] specializing in [domain].
When invoked:
1. [Step 1]
2. [Step 2]
3. [Step 3]
[Detailed instructions]
[Checklists]
[Best practices]
[Output format]
Integration Workflows
Workflow 1: Feature Development
1. product-planner → Creates requirements
2. frontend-developer + backend-developer → Build (parallel)
3. test-runner → Validates (sequential)
4. code-reviewer → Reviews (sequential)
Workflow 2: Bug Fix
1. debugger → Analyzes issue
2. [appropriate-dev-agent] → Fixes
3. test-runner → Validates fix
Workflow 3: Code Review
1. code-reviewer → Quality review (can run solo)
2. security-auditor → Security scan (can run solo)
MCP Tool Integration
Common MCP servers to integrate:
- mcp__github: PR reviews, issues, repo operations
- mcp__playwright: E2E testing, screenshots, browser automation
- mcp__context7: Documentation search, knowledge queries
- mcp__filesystem: Advanced file operations
- Custom MCP servers: Any user-configured MCP tools
Agents automatically reference MCP tools in their capabilities when configured.
Safety & Performance
Process Monitoring
Agents consume system resources. Monitor with:
ps aux | grep -E "mcp|npm|claude" | wc -l
Safe ranges:
- 15-20: Strategic agents (parallel)
- 20-30: Implementation agents (coordinated)
- 12-18: Quality agents (sequential)
Warnings:
30: Reduce parallelization
60: Critical - restart system
Execution Rules
✅ Safe: 4-5 strategic agents in parallel ✅ Safe: 2-3 implementation agents coordinated ❌ Unsafe: Quality agents in parallel (crashes system)
Best Practices
- Keep agents focused - One clear responsibility per agent
- Use descriptive descriptions - Enables auto-invocation
- Follow tool access patterns - Match tools to agent type
- Specify execution pattern - Prevents performance issues
- Leverage MCP tools - Enhance agent capabilities
- Test agents incrementally - Start simple, add complexity
- Version control agents - Check project agents into git
Limitations
- Agents are templates - customize for your specific needs
- Tool suggestions are guidelines, not requirements
- MCP tools require servers to be configured
- Performance depends on system resources
- Generated agents need testing in your environment
Installation
Generated Agent Files:
Place in one of these locations:
Project agents (shared with team):
.claude/agents/custom-agent.md
Personal agents (available everywhere):
~/.claude/agents/custom-agent.md
When to Use This Skill
Create custom agents for:
- Domain-specific workflows (data science, ML, finance)
- Team-specific conventions (your code style, testing approach)
- Specialized tools or frameworks (Shopify, AWS, Kubernetes)
- Custom MCP server integrations
- Rapid prototyping of agent ideas
Use the AGENTS_FACTORY_PROMPT.md template when:
- You need multiple related agents
- You want consistent agent patterns
- You're building an agentic framework
- You want to test agent concepts quickly
Version: 1.0.0 Last Updated: October 22, 2025 Compatibility: Claude Code (agents system) Template Location: documentation/templates/AGENTS_FACTORY_PROMPT.md
Source: ComeOnOliver/skillshub — distributed by TomeVault.
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核