Agent 模板

LibreFang 预装了 30 个 Agent 模板，开箱即用。

30 个内置模板

Agent	描述	能力
hello-world	入门示例，可读取文件、搜索网页	file_read, file_list, web_fetch, web_search
coder	代码开发和调试	code_execute, file_read, file_write, git
researcher	深度研究和分析	web_search, web_fetch, memory_recall
writer	内容创作和编辑	file_write, memory_recall
analyst	数据分析和洞察	data_analysis, file_read
architect	系统架构设计	memory_recall, web_fetch
ops	运维自动化	shell_execute, file_read
security-auditor	安全审计	code_analysis, security_scan
planner	项目规划和任务管理	memory_store, memory_recall
debugger	问题诊断和调试	code_analysis, shell_execute
translator	多语言翻译	memory_recall
tutor	学习和辅导	memory_recall, web_fetch
personal-finance	个人财务管理	data_analysis, file_read
travel-planner	旅行规划	web_search, memory_recall
social-media	社交媒体管理	web_fetch, memory_store
meeting-assistant	会议助手	calendar, email
email-assistant	邮件助手	email, file_read
recruiter	招聘和人才发现	web_search, memory_recall
data-scientist	数据科学分析	python, pandas, jupyter
test-engineer	测试工程	code_execute, file_read
doc-writer	文档编写	file_write, memory_recall
customer-support	客户支持	memory_recall, email
sales-assistant	销售助手	web_search, memory_store
legal-assistant	法律助手	web_fetch, memory_recall
health-tracker	健康追踪	data_analysis, memory_store
home-automation	家居自动化	shell_execute, web_fetch
assistant	通用助手	web_fetch, memory_recall
orchestrator	多 Agent 协调	agent_spawn
devops-lead	DevOps 领导	shell_execute, kubernetes

使用模板

生成 Agent

librefang agent spawn agents/hello-world/agent.toml

查看模板

ls agents/

模板结构

agents/
└── hello-world/
    ├── agent.toml      # Agent 清单
    └── SKILL.md      # (可选) 领域知识

agent.toml 示例

name = "hello-world"
version = "0.1.0"
description = "A friendly greeting agent"
author = "librefang"
module = "builtin:chat"

[model]
provider = "default"
model = "default"
max_tokens = 4096
temperature = 0.6

[system_prompt]
prompt = "You are Hello World..."

[capabilities]
tools = ["file_read", "file_list", "web_fetch"]
network = ["*"]
memory_read = ["*"]
memory_write = ["self.*"]

[resources]
max_llm_tokens_per_hour = 100000

创建自定义 Agent

最小配置

name = "my-agent"
version = "0.1.0"
description = "My custom agent"
module = "builtin:chat"

[model]
provider = "groq"
model = "llama-3.3-70b-versatile"

完整配置

name = "my-agent"
version = "0.1.0"
description = "A helpful assistant"
author = "you"
module = "builtin:chat"

[model]
provider = "default"
model = "default"
max_tokens = 4096
temperature = 0.7

[system_prompt]
prompt = "You are a helpful assistant..."

[resources]
max_llm_tokens_per_hour = 100000
max_iterations = 100

[capabilities]
tools = ["file_read", "file_list", "web_fetch"]
network = ["*"]
memory_read = ["*"]
memory_write = ["self.*"]
agent_spawn = false

[schedule]
enabled = false
cron = "0 9 * * *"

[channels]
telegram_enabled = true
discord_enabled = false

Agent 字段参考

字段	类型	说明
`name`	String	Agent 名称
`version`	String	版本号
`description`	String	描述
`author`	String	作者
`module`	String	模块 (builtin:chat)
`model.provider`	String	LLM 提供商
`model.model`	String	模型名称
`model.max_tokens`	Integer	最大 token
`model.temperature`	Float	采样温度
`system_prompt`	String	系统提示
`capabilities.tools`	Array	可用工具
`capabilities.network`	Array	网络访问
`capabilities.memory_read`	Array	记忆读取
`capabilities.memory_write`	Array	记忆写入
`capabilities.agent_spawn`	Boolean	是否可以生成子 Agent

工具列表

工具	说明
file_read	读取文件
file_write	写入文件
file_list	列出目录
shell_execute	执行 Shell
web_fetch	获取网页
web_search	搜索网页
memory_store	存储记忆
memory_recall	回忆记忆
code_execute	执行代码
git	Git 操作
python	Python REPL
jupyter	Jupyter 笔记本

模块类型

模块	说明
builtin:chat	通用聊天
builtin:researcher	研究
builtin:writer	写作
builtin:coder	编码

CLI 命令

# 生成 Agent
librefang agent spawn agents/hello-world/agent.toml

# 列出所有 Agent
librefang agent list

# 与 Agent 聊天
librefang agent chat <agent-id>

# 终止 Agent
librefang agent kill <agent-id>

# 查看 Agent 状态
librefang agent status <agent-id>