Hermes Agent Framework

Skill by ara.so — Hermes Skills collection.

Hermes Agent is an open-source AI Agent framework by Nous Research that features a built-in self-improving learning loop, three-layer memory system (episodic, semantic, procedural), and automatic Skill creation and evolution. Unlike traditional agentic frameworks, Hermes continuously learns from interactions and builds up capabilities over time.

由ara.so提供的Skill——Hermes技能集合。

Hermes Agent是Nous Research开发的开源AI Agent框架，具备内置自改进学习循环、三层记忆系统（episodic, semantic, procedural）以及Skill自动创建与进化功能。与传统智能体框架不同，Hermes能从交互中持续学习，逐步提升自身能力。

Installation

安装

Prerequisites

前置条件

Python 3.9+
API key for LLM provider (OpenAI, Anthropic, etc.)

Python 3.9+
LLM提供商的API密钥（OpenAI、Anthropic等）

Basic Installation

基础安装

bash

undefined

bash

undefined

Clone the repository

git clone https://github.com/NousResearch/hermes-agent.git cd hermes-agent

Install dependencies

pip install -r requirements.txt

Or install via pip (if published)

pip install hermes-agent

undefined

pip install hermes-agent

undefined

Configuration

配置

Create a

.env

file in the project root:

bash

undefined

在项目根目录创建

.env

文件：

bash

undefined

LLM Provider Configuration

OPENAI_API_KEY=your_openai_key_here ANTHROPIC_API_KEY=your_anthropic_key_here

Agent Configuration

HERMES_MODEL=gpt-4 HERMES_MEMORY_PATH=./memory HERMES_SKILLS_PATH=./skills

undefined

HERMES_MODEL=gpt-4 HERMES_MEMORY_PATH=./memory HERMES_SKILLS_PATH=./skills

undefined

Core Concepts

核心概念

Three-Layer Memory System

三层记忆系统

Episodic Memory: Stores conversation history and interaction sequences
Semantic Memory: Long-term knowledge and facts extracted from experiences
Procedural Memory: Skills and learned procedures (how to do things)

Episodic Memory：存储对话历史和交互序列
Semantic Memory：从经验中提取的长期知识与事实
Procedural Memory：技能与习得的操作流程（如何完成任务）

Learning Loop

学习循环

Hermes operates in a continuous cycle:

Perceive: Receive user input and context
Reflect: Analyze what happened and extract learnings
Learn: Update memory systems and create/modify Skills
Act: Execute tasks using available tools and Skills

Hermes运行在持续循环中：

感知：接收用户输入和上下文
反思：分析事件过程并提取经验
学习：更新记忆系统并创建/修改Skill
行动：使用可用工具和Skill执行任务

Skills

Skills are reusable capabilities that Hermes creates and refines automatically. They're stored as structured modules in the procedural memory.

Skills是Hermes自动创建和优化的可复用能力，以结构化模块形式存储在过程记忆中。

Basic Usage

基础用法

Starting a Hermes Agent

启动Hermes Agent

python

from hermes_agent import HermesAgent, Config

python

from hermes_agent import HermesAgent, Config

Initialize configuration

config = Config( model="gpt-4", memory_path="./memory", skills_path="./skills", temperature=0.7 )

Create agent instance

agent = HermesAgent(config)

Start conversation

response = agent.chat("Help me analyze this CSV file and create visualizations") print(response)

undefined

response = agent.chat("Help me analyze this CSV file and create visualizations") print(response)

undefined

With Custom System Prompt

使用自定义系统提示词

python

from hermes_agent import HermesAgent, Config

config = Config(
    model="claude-3-5-sonnet-20241022",
    system_prompt="""You are a specialized data analysis agent.
    Focus on statistical rigor and clear visualizations.
    Always explain your analytical choices."""
)

agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config

config = Config(
    model="claude-3-5-sonnet-20241022",
    system_prompt="""You are a specialized data analysis agent.
    Focus on statistical rigor and clear visualizations.
    Always explain your analytical choices."""
)

agent = HermesAgent(config)

Enabling Memory Persistence

启用记忆持久化

python

from hermes_agent import HermesAgent, Config, MemoryConfig

memory_config = MemoryConfig(
    episodic_enabled=True,
    semantic_enabled=True,
    procedural_enabled=True,
    retention_days=90,
    auto_consolidate=True
)

config = Config(
    model="gpt-4",
    memory_config=memory_config
)

agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config, MemoryConfig

memory_config = MemoryConfig(
    episodic_enabled=True,
    semantic_enabled=True,
    procedural_enabled=True,
    retention_days=90,
    auto_consolidate=True
)

config = Config(
    model="gpt-4",
    memory_config=memory_config
)

agent = HermesAgent(config)

Memory is automatically saved and loaded

agent.chat("Remember that I prefer Python over JavaScript")

Later sessions will recall this preference

undefined

undefined

Working with Skills

技能操作

Creating a Custom Skill

创建自定义Skill

python

from hermes_agent import Skill, SkillParameter

python

from hermes_agent import Skill, SkillParameter

Define a custom skill

web_scraper_skill = Skill( name="web_scraper", description="Scrape and extract structured data from websites", parameters=[ SkillParameter(name="url", type="string", required=True), SkillParameter(name="selectors", type="object", required=False) ], implementation=""" import requests from bs4 import BeautifulSoup

def execute(url, selectors=None): response = requests.get(url) soup = BeautifulSoup(response.content, 'html.parser')

if selectors:
    results = {}
    for key, selector in selectors.items():
        results[key] = soup.select(selector)
    return results

return soup.get_text()

""" )

web_scraper_skill = Skill( name="web_scraper", description="Scrape and extract structured data from websites", parameters=[ SkillParameter(name="url", type="string", required=True), SkillParameter(name="selectors", type="object", required=False) ], implementation=""" import requests from bs4 import BeautifulSoup

def execute(url, selectors=None): response = requests.get(url) soup = BeautifulSoup(response.content, 'html.parser')

if selectors:
    results = {}
    for key, selector in selectors.items():
        results[key] = soup.select(selector)
    return results

return soup.get_text()

""" )

Register skill with agent

agent.register_skill(web_scraper_skill)

undefined

agent.register_skill(web_scraper_skill)

undefined

Loading Skills from Directory

从目录加载Skill

python

from hermes_agent import HermesAgent, Config

config = Config(
    model="gpt-4",
    skills_path="./my_custom_skills"
)

agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config

config = Config(
    model="gpt-4",
    skills_path="./my_custom_skills"
)

agent = HermesAgent(config)

Agent automatically loads all skills from directory

Skills are available for use in conversations

undefined

undefined

Skill Auto-Evolution

技能自动进化

python

from hermes_agent import HermesAgent, Config, LearningConfig

learning_config = LearningConfig(
    auto_create_skills=True,
    skill_refinement=True,
    min_usage_for_creation=3  # Create skill after pattern used 3+ times
)

config = Config(
    model="gpt-4",
    learning_config=learning_config
)

agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config, LearningConfig

learning_config = LearningConfig(
    auto_create_skills=True,
    skill_refinement=True,
    min_usage_for_creation=3  # Create skill after pattern used 3+ times
)

config = Config(
    model="gpt-4",
    learning_config=learning_config
)

agent = HermesAgent(config)

As agent performs repeated tasks, it automatically creates reusable skills

agent.chat("Convert this JSON to CSV format") agent.chat("Convert this other JSON to CSV") agent.chat("And convert this JSON to CSV too")

After 3rd usage, Hermes creates a "json_to_csv" skill automatically

undefined

undefined

Tool Integration

工具集成

Registering External Tools

注册外部工具

python

from hermes_agent import HermesAgent, Tool

def search_api(query: str) -> dict:
    """Search using external API"""
    import os
    import requests
    
    api_key = os.getenv("SEARCH_API_KEY")
    response = requests.get(
        "https://api.example.com/search",
        params={"q": query, "key": api_key}
    )
    return response.json()

python

from hermes_agent import HermesAgent, Tool

def search_api(query: str) -> dict:
    """Search using external API"""
    import os
    import requests
    
    api_key = os.getenv("SEARCH_API_KEY")
    response = requests.get(
        "https://api.example.com/search",
        params={"q": query, "key": api_key}
    )
    return response.json()

Register as tool

search_tool = Tool( name="web_search", description="Search the web for current information", function=search_api, parameters={ "query": {"type": "string", "description": "Search query"} } )

agent = HermesAgent(config) agent.register_tool(search_tool)

undefined

search_tool = Tool( name="web_search", description="Search the web for current information", function=search_api, parameters={ "query": {"type": "string", "description": "Search query"} } )

agent = HermesAgent(config) agent.register_tool(search_tool)

undefined

Built-in Tool Categories

内置工具类别

python

from hermes_agent import HermesAgent, Config, ToolConfig

tool_config = ToolConfig(
    enable_file_operations=True,
    enable_web_browsing=True,
    enable_code_execution=True,
    enable_shell_commands=False,  # Disabled by default for security
    allowed_domains=["*.example.com", "api.trusted.com"]
)

config = Config(
    model="gpt-4",
    tool_config=tool_config
)

agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config, ToolConfig

tool_config = ToolConfig(
    enable_file_operations=True,
    enable_web_browsing=True,
    enable_code_execution=True,
    enable_shell_commands=False,  # Disabled by default for security
    allowed_domains=["*.example.com", "api.trusted.com"]
)

config = Config(
    model="gpt-4",
    tool_config=tool_config
)

agent = HermesAgent(config)

Multi-Agent Orchestration

多Agent编排

Creating Agent Teams

创建Agent团队

python

from hermes_agent import HermesAgent, AgentTeam, Config

python

from hermes_agent import HermesAgent, AgentTeam, Config

Create specialized agents

researcher = HermesAgent(Config( model="gpt-4", system_prompt="You are a research specialist. Focus on gathering and analyzing information." ))

coder = HermesAgent(Config( model="claude-3-5-sonnet-20241022", system_prompt="You are a coding specialist. Write clean, efficient code." ))

writer = HermesAgent(Config( model="gpt-4", system_prompt="You are a technical writer. Create clear documentation." ))

researcher = HermesAgent(Config( model="gpt-4", system_prompt="You are a research specialist. Focus on gathering and analyzing information." ))

coder = HermesAgent(Config( model="claude-3-5-sonnet-20241022", system_prompt="You are a coding specialist. Write clean, efficient code." ))

writer = HermesAgent(Config( model="gpt-4", system_prompt="You are a technical writer. Create clear documentation." ))

Create team

team = AgentTeam( agents=[researcher, coder, writer], coordinator=HermesAgent(Config( model="gpt-4", system_prompt="Coordinate agent activities and synthesize results." )) )

Execute team task

result = team.execute( "Research best practices for API design, implement a sample API, and document it" )

undefined

result = team.execute( "Research best practices for API design, implement a sample API, and document it" )

undefined

Agent Communication

Agent通信

python

from hermes_agent import HermesAgent, AgentChannel

python

from hermes_agent import HermesAgent, AgentChannel

Create communication channel

channel = AgentChannel()

agent_a = HermesAgent(config) agent_b = HermesAgent(config)

channel = AgentChannel()

agent_a = HermesAgent(config) agent_b = HermesAgent(config)

Connect agents to channel

agent_a.connect(channel) agent_b.connect(channel)

Agents can now share context and learnings

agent_a.chat("Learn about Python async patterns")

agent_b automatically has access to what agent_a learned

agent_b.chat("Use async patterns to build a web scraper")

undefined

agent_b.chat("Use async patterns to build a web scraper")

undefined

Advanced Configuration

高级配置

Feedback Loop Customization

反馈循环自定义

python

from hermes_agent import HermesAgent, Config, FeedbackConfig

feedback_config = FeedbackConfig(
    enable_self_critique=True,
    reflection_frequency="after_task",  # or "periodic", "never"
    quality_threshold=0.8,
    auto_correction=True
)

config = Config(
    model="gpt-4",
    feedback_config=feedback_config
)

agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config, FeedbackConfig

feedback_config = FeedbackConfig(
    enable_self_critique=True,
    reflection_frequency="after_task",  # or "periodic", "never"
    quality_threshold=0.8,
    auto_correction=True
)

config = Config(
    model="gpt-4",
    feedback_config=feedback_config
)

agent = HermesAgent(config)

Constraints and Safety

约束与安全

python

from hermes_agent import HermesAgent, Config, ConstraintConfig

constraints = ConstraintConfig(
    max_iterations=10,
    timeout_seconds=300,
    max_tool_calls_per_turn=5,
    blocked_operations=["rm -rf", "DROP TABLE"],
    require_approval_for=["file_delete", "api_payment"]
)

config = Config(
    model="gpt-4",
    constraint_config=constraints
)

agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config, ConstraintConfig

constraints = ConstraintConfig(
    max_iterations=10,
    timeout_seconds=300,
    max_tool_calls_per_turn=5,
    blocked_operations=["rm -rf", "DROP TABLE"],
    require_approval_for=["file_delete", "api_payment"]
)

config = Config(
    model="gpt-4",
    constraint_config=constraints
)

agent = HermesAgent(config)

Memory Management

内存管理

python

from hermes_agent import HermesAgent, Config

config = Config(model="gpt-4")
agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config

config = Config(model="gpt-4")
agent = HermesAgent(config)

Inspect memory

episodic = agent.memory.get_episodic(last_n=10) semantic = agent.memory.get_semantic(topic="python programming") skills = agent.memory.get_skills()

Clear specific memory types

agent.memory.clear_episodic() # Clear conversation history agent.memory.clear_semantic(topic="outdated_info")

Export/Import memory

agent.memory.export("backup.json") agent.memory.import_from("backup.json")

undefined

agent.memory.export("backup.json") agent.memory.import_from("backup.json")

undefined

Real-World Examples

实际应用示例

Personal Knowledge Assistant

个人知识助手

python

from hermes_agent import HermesAgent, Config, MemoryConfig, ToolConfig

memory_config = MemoryConfig(
    episodic_enabled=True,
    semantic_enabled=True,
    retention_days=365,
    auto_consolidate=True
)

tool_config = ToolConfig(
    enable_file_operations=True,
    enable_web_browsing=True
)

config = Config(
    model="gpt-4",
    memory_config=memory_config,
    tool_config=tool_config,
    system_prompt="""You are a personal knowledge assistant.
    Learn from all our interactions and help me recall information,
    make connections, and build on past conversations."""
)

agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config, MemoryConfig, ToolConfig

memory_config = MemoryConfig(
    episodic_enabled=True,
    semantic_enabled=True,
    retention_days=365,
    auto_consolidate=True
)

tool_config = ToolConfig(
    enable_file_operations=True,
    enable_web_browsing=True
)

config = Config(
    model="gpt-4",
    memory_config=memory_config,
    tool_config=tool_config,
    system_prompt="""You are a personal knowledge assistant.
    Learn from all our interactions and help me recall information,
    make connections, and build on past conversations."""
)

agent = HermesAgent(config)

Over time, agent builds up knowledge about user preferences, projects, etc.

agent.chat("I'm working on a new Python project for data analysis")

Days later...

agent.chat("What was that project I mentioned last week?")

undefined

agent.chat("What was that project I mentioned last week?")

undefined

Development Automation Agent

开发自动化Agent

python

from hermes_agent import HermesAgent, Config, ToolConfig, LearningConfig

tool_config = ToolConfig(
    enable_code_execution=True,
    enable_file_operations=True,
    enable_shell_commands=True
)

learning_config = LearningConfig(
    auto_create_skills=True,
    skill_refinement=True
)

config = Config(
    model="claude-3-5-sonnet-20241022",
    tool_config=tool_config,
    learning_config=learning_config,
    system_prompt="You are a development automation specialist."
)

agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config, ToolConfig, LearningConfig

tool_config = ToolConfig(
    enable_code_execution=True,
    enable_file_operations=True,
    enable_shell_commands=True
)

learning_config = LearningConfig(
    auto_create_skills=True,
    skill_refinement=True
)

config = Config(
    model="claude-3-5-sonnet-20241022",
    tool_config=tool_config,
    learning_config=learning_config,
    system_prompt="You are a development automation specialist."
)

agent = HermesAgent(config)

Agent learns common development patterns and creates skills

agent.chat("Set up a new FastAPI project with PostgreSQL") agent.chat("Add authentication with JWT") agent.chat("Create CRUD endpoints for a User model")

Agent creates reusable skills for these common patterns

undefined

undefined

Content Creation Pipeline

内容创作流水线

python

from hermes_agent import HermesAgent, AgentTeam, Config

researcher = HermesAgent(Config(
    model="gpt-4",
    system_prompt="Research topics and gather information.",
    tool_config=ToolConfig(enable_web_browsing=True)
))

writer = HermesAgent(Config(
    model="claude-3-5-sonnet-20241022",
    system_prompt="Create engaging, well-structured content."
))

editor = HermesAgent(Config(
    model="gpt-4",
    system_prompt="Review and refine content for clarity and quality."
))

team = AgentTeam(agents=[researcher, writer, editor])

python

from hermes_agent import HermesAgent, AgentTeam, Config

researcher = HermesAgent(Config(
    model="gpt-4",
    system_prompt="Research topics and gather information.",
    tool_config=ToolConfig(enable_web_browsing=True)
))

writer = HermesAgent(Config(
    model="claude-3-5-sonnet-20241022",
    system_prompt="Create engaging, well-structured content."
))

editor = HermesAgent(Config(
    model="gpt-4",
    system_prompt="Review and refine content for clarity and quality."
))

team = AgentTeam(agents=[researcher, writer, editor])

Automated content pipeline

result = team.execute( "Create a comprehensive blog post about Hermes Agent framework" )

undefined

result = team.execute( "Create a comprehensive blog post about Hermes Agent framework" )

undefined

CLI Usage

CLI使用

If Hermes provides a command-line interface:

bash

undefined

如果Hermes提供命令行界面：

bash

undefined

Start interactive session

hermes chat

With specific model

hermes chat --model gpt-4

Load skills from directory

hermes chat --skills ./my_skills

Enable debug mode

hermes chat --debug

One-off command

hermes exec "analyze this CSV: data.csv"

Manage memory

hermes memory export backup.json hermes memory import backup.json hermes memory clear --episodic

List learned skills

hermes skills list

Export a skill

hermes skills export web_scraper > web_scraper.py

undefined

hermes skills export web_scraper > web_scraper.py

undefined

Troubleshooting

故障排除

Memory Not Persisting

记忆未持久化

python

undefined

python

undefined

Ensure memory path is writable

import os from hermes_agent import HermesAgent, Config

memory_path = "./hermes_memory" os.makedirs(memory_path, exist_ok=True)

config = Config( model="gpt-4", memory_path=memory_path, auto_save=True # Enable automatic saving )

agent = HermesAgent(config)

undefined

import os from hermes_agent import HermesAgent, Config

memory_path = "./hermes_memory" os.makedirs(memory_path, exist_ok=True)

config = Config( model="gpt-4", memory_path=memory_path, auto_save=True # Enable automatic saving )

agent = HermesAgent(config)

undefined

Skills Not Loading

技能未加载

python

undefined

python

undefined

Verify skills directory structure

from hermes_agent import HermesAgent, Config

config = Config( model="gpt-4", skills_path="./skills", debug=True # Enable debug logging )

agent = HermesAgent(config)

from hermes_agent import HermesAgent, Config

config = Config( model="gpt-4", skills_path="./skills", debug=True # Enable debug logging )

agent = HermesAgent(config)

Check loaded skills

print(agent.list_skills())

undefined

print(agent.list_skills())

undefined

High Token Usage

Token使用量过高

python

from hermes_agent import HermesAgent, Config, MemoryConfig

python

from hermes_agent import HermesAgent, Config, MemoryConfig

Optimize memory retrieval

memory_config = MemoryConfig( max_episodic_context=5, # Limit conversation history semantic_relevance_threshold=0.7, # Only retrieve relevant memories consolidation_frequency="daily" # Compress old memories )

config = Config( model="gpt-4", memory_config=memory_config, max_tokens=2000 # Limit response length )

agent = HermesAgent(config)

undefined

memory_config = MemoryConfig( max_episodic_context=5, # Limit conversation history semantic_relevance_threshold=0.7, # Only retrieve relevant memories consolidation_frequency="daily" # Compress old memories )

config = Config( model="gpt-4", memory_config=memory_config, max_tokens=2000 # Limit response length )

agent = HermesAgent(config)

undefined

Tool Execution Failures

工具执行失败

python

from hermes_agent import HermesAgent, Config, ToolConfig

tool_config = ToolConfig(
    timeout_seconds=30,
    retry_attempts=3,
    error_handling="graceful",  # vs "strict"
    log_tool_calls=True
)

config = Config(
    model="gpt-4",
    tool_config=tool_config
)

agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config, ToolConfig

tool_config = ToolConfig(
    timeout_seconds=30,
    retry_attempts=3,
    error_handling="graceful",  # vs "strict"
    log_tool_calls=True
)

config = Config(
    model="gpt-4",
    tool_config=tool_config
)

agent = HermesAgent(config)

Rate Limiting Issues

速率限制问题

python

from hermes_agent import HermesAgent, Config

config = Config(
    model="gpt-4",
    rate_limit_rpm=20,  # Requests per minute
    backoff_strategy="exponential",
    retry_on_rate_limit=True
)

agent = HermesAgent(config)

python

from hermes_agent import HermesAgent, Config

config = Config(
    model="gpt-4",
    rate_limit_rpm=20,  # Requests per minute
    backoff_strategy="exponential",
    retry_on_rate_limit=True
)

agent = HermesAgent(config)

Best Practices

最佳实践

Start Simple: Begin with basic configuration and add complexity as needed
Enable Memory: Hermes's strength is learning over time - enable all memory systems
Curate Skills: Review auto-created skills periodically and refine them
Set Constraints: Always configure safety constraints for production use
Monitor Token Usage: Use memory consolidation to manage costs
Version Skills: Export and version control important skills
Use Teams Wisely: Specialized agents work better than one generalist for complex tasks
Provide Feedback: The more feedback in the loop, the better Hermes learns

从简开始：先使用基础配置，再根据需求增加复杂度
启用记忆：Hermes的优势在于长期学习——启用所有记忆系统
整理技能：定期审核自动创建的技能并优化
设置约束：生产环境中务必配置安全约束
监控Token使用：通过记忆整合控制成本
版本化技能：导出并版本控制重要技能
合理使用团队：处理复杂任务时，专用Agent比通用Agent效果更好
提供反馈：反馈循环越完善，Hermes的学习效果越好

Resources

资源

Official Documentation: https://hermes-agent.nousresearch.com/docs/
GitHub Repository: https://github.com/NousResearch/hermes-agent
Orange Book Guide: https://github.com/alchaincyf/hermes-agent-orange-book
Community Discord: Check repository for link
Nous Research: https://nousresearch.com/

官方文档：https://hermes-agent.nousresearch.com/docs/
GitHub仓库：https://github.com/NousResearch/hermes-agent
橙皮书指南：https://github.com/alchaincyf/hermes-agent-orange-book
社区Discord：查看仓库获取链接
Nous Research：https://nousresearch.com/

hermes-agent-framework

Original

Translation

Hermes Agent Framework

Hermes Agent Framework

Installation

安装

Prerequisites

前置条件

Basic Installation

基础安装

Clone the repository

Clone the repository

Install dependencies

Install dependencies

Or install via pip (if published)

Or install via pip (if published)

Configuration

配置

LLM Provider Configuration

LLM Provider Configuration

Agent Configuration

Agent Configuration

Core Concepts

核心概念

Three-Layer Memory System

三层记忆系统

Learning Loop

学习循环

Skills

Skills

Basic Usage

基础用法

Starting a Hermes Agent

启动Hermes Agent

Initialize configuration

Initialize configuration

Create agent instance

Create agent instance

Start conversation

Start conversation

With Custom System Prompt

使用自定义系统提示词

Enabling Memory Persistence

启用记忆持久化

Memory is automatically saved and loaded

Memory is automatically saved and loaded

Later sessions will recall this preference

Later sessions will recall this preference

Working with Skills

技能操作

Creating a Custom Skill

创建自定义Skill

Define a custom skill

Define a custom skill

Register skill with agent

Register skill with agent

Loading Skills from Directory

从目录加载Skill

Agent automatically loads all skills from directory

Agent automatically loads all skills from directory

Skills are available for use in conversations

Skills are available for use in conversations

Skill Auto-Evolution

技能自动进化

As agent performs repeated tasks, it automatically creates reusable skills

As agent performs repeated tasks, it automatically creates reusable skills

After 3rd usage, Hermes creates a "json_to_csv" skill automatically

After 3rd usage, Hermes creates a "json_to_csv" skill automatically

Tool Integration

工具集成

Registering External Tools

注册外部工具

Register as tool

Register as tool

Built-in Tool Categories

内置工具类别

Multi-Agent Orchestration

多Agent编排

Creating Agent Teams