liusongtao
/
base_agent
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

in devel

This commit is contained in:
liusongtao 2026-04-15 16:20:22 +08:00
parent f184c199c9
commit 38dcc420b4
16 changed files with 363 additions and 177 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
agent/agent.py
View File

@ -399,32 +399,11 @@ def create_agent() -> tuple[Agent, SkillRegistry]:
Returns: Returns:
(agent, registry) —— registry 需在程序退出时调用 .close() (agent, registry) —— registry 需在程序退出时调用 .close()
""" """
from tools.calculator import CalculatorTool
from tools.code_executor import CodeExecutorTool
from tools.file_reader import FileReaderTool
from tools.ssh_docker import SSHDockerTool
from tools.static_analyzer import StaticAnalyzerTool
from tools.web_search import WebSearchTool
registry = SkillRegistry() registry = SkillRegistry()
registry.load_local_skills(settings.skills_directory)
# 注册本地工具（根据 config.yaml mcp.enabled_tools 过滤） registry.load_local_tools()
enabled = settings.mcp.enabled_tools
tool_map = {
"calculator": CalculatorTool,
"web_search": WebSearchTool,
"file_reader": FileReaderTool,
"code_executor": CodeExecutorTool,
"static_analyzer": StaticAnalyzerTool,
"ssh_docker": SSHDockerTool,
}
for name in enabled:
if name in tool_map:
registry.register_local(tool_map[name]())
# 连接在线 MCP Skill（来自 config.yaml mcp_skills） # 连接在线 MCP Skill（来自 config.yaml mcp_skills）
registry.connect_skills() registry.connect_mcp_skills()
agent = Agent(registry) agent = Agent(registry)
return agent, registry return agent, registry

3
config.yaml
View File

@ -16,8 +16,7 @@ llm:
stream: false stream: false
model_path: "" model_path: ""
ollama_host: "http://localhost:11434" ollama_host: "http://localhost:11434"
skills_directory: "C:\\Users\\sonto\\Workspace\\base_agent\\skills\\OpenCLI\\skills" # 新增：SKILL.md 文件所在目录
skills_directory: "./skills" # 新增：SKILL.md 文件所在目录
# ── 本地 MCP Server 配置 ─────────────────────────────────────── # ── 本地 MCP Server 配置 ───────────────────────────────────────
mcp: mcp:
server_name: "DemoMCPServer" server_name: "DemoMCPServer"

2
config/settings.py
View File

@ -375,7 +375,7 @@ class ConfigLoader:
memory=cls._build_memory(raw.get("memory", {})), memory=cls._build_memory(raw.get("memory", {})),
logging=cls._build_logging(raw.get("logging", {})), logging=cls._build_logging(raw.get("logging", {})),
agent=cls._build_agent(raw.get("agent", {})), agent=cls._build_agent(raw.get("agent", {})),
skills=raw.get("skills", []) skills_directory=raw.get("skills_directory", "")
) )
# ── LLM ─────────────────────────────────────────────────── # ── LLM ───────────────────────────────────────────────────

6
llm/providers/openai_provider.py
View File

@ -70,12 +70,6 @@ class OpenAIProvider(BaseProvider):
3. 当后续步骤依赖前步结果时，先完成前步再继续 3. 当后续步骤依赖前步结果时，先完成前步再继续
4. 每次只规划并调用当前最合适的工具 4. 每次只规划并调用当前最合适的工具
5. 所有工具执行完毕后，整合结果给出最终回复 5. 所有工具执行完毕后，整合结果给出最终回复
## 重要规则
- 数学计算必须使用 calculator 工具，不要自行计算
- 需要实时信息时使用 web_search 工具
- 文件操作使用 file_reader 工具
- 代码执行使用 code_executor 工具
""" """
# 回复生成系统 Prompt # 回复生成系统 Prompt

211
logs/agent.log
View File

@ -6140,3 +6140,214 @@ The function `get_system_name()` uses `platform.system()` to determine the syste
重试 : 2 次 重试 : 2 次
2026-04-13 16:12:50 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999 2026-04-13 16:12:50 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-13 16:12:50 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=48433 2026-04-13 16:12:50 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=48433
2026-04-15 15:34:11 [WARNING ] agent.TOOL.SSHDocker │ ⚠️ paramiko 未安装，请执行: pip install paramiko>=3.0.0
2026-04-15 15:34:11 [WARNING ] agent.mcp.SkillLoader │ 未在opencli-autofix发现SKILL.md文件
2026-04-15 15:34:11 [WARNING ] agent.mcp.SkillLoader │ 未在opencli-browser发现SKILL.md文件
2026-04-15 15:34:11 [WARNING ] agent.mcp.SkillLoader │ 未在opencli-explorer发现SKILL.md文件
2026-04-15 15:34:11 [WARNING ] agent.mcp.SkillLoader │ 未在opencli-oneshot发现SKILL.md文件
2026-04-15 15:34:11 [WARNING ] agent.mcp.SkillLoader │ 未在opencli-usage发现SKILL.md文件
2026-04-15 15:34:11 [WARNING ] agent.mcp.SkillLoader │ 未在smart-search发现SKILL.md文件
2026-04-15 15:34:11 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: calculator
2026-04-15 15:34:11 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: web_search
2026-04-15 15:34:11 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: file_reader
2026-04-15 15:34:11 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: code_executor
2026-04-15 15:34:11 [INFO ] agent.MCP.SkillRegistry │ 🌐 开始连接在线 MCP Skills，数量=1
2026-04-15 15:34:11 [INFO ] agent.MCP.SkillClient │ 🌐 连接在线 MCP Skill: [hexstrike-ai]
传输协议: stdio
地址 : python3
超时 : 300s
重试 : 2 次
2026-04-15 15:34:11 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 15:34:11 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=19704
2026-04-15 15:34:11 [WARNING ] agent.MCP.SkillClient │ ⚠️ 连接失败 (attempt 1/3)，1s 后重试: stdio 子进程无响应 skill=hexstrike-ai method=initialize
2026-04-15 15:34:12 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 15:34:12 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 15:34:13 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=14668
2026-04-15 15:34:13 [WARNING ] agent.MCP.SkillClient │ ⚠️ 连接失败 (attempt 2/3)，2s 后重试: stdio 子进程无响应 skill=hexstrike-ai method=initialize
2026-04-15 15:34:15 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 15:34:15 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 15:34:15 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=13920
2026-04-15 15:34:15 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 15:34:15 [ERROR ] agent.MCP.SkillRegistry │ ❌ Skill [hexstrike-ai] 连接失败，跳过
错误: ❌ MCP Skill [hexstrike-ai] 连接失败（已重试 2 次）
最后错误: stdio 子进程无响应 skill=hexstrike-ai method=initialize
2026-04-15 15:34:15 [DEBUG ] agent.MCP.SkillClient │ 🔌 MCP Skill [hexstrike-ai] 已断开
2026-04-15 15:34:15 [INFO ] agent.MCP.SkillRegistry │ 📊 SkillRegistry 初始化完成
本地工具 : 4 个 ['calculator', 'web_search', 'file_reader', 'code_executor']
远端工具 : 0 个 []
2026-04-15 15:34:16 [INFO ] agent.Agent │ 🤖 Agent 初始化完成
LLM : openai / gpt-4o
工具总数 : 4 个
最大步数 : 10
工具列表 :
🔵 [local ] calculator
🔵 [local ] web_search
🔵 [local ] file_reader
🔵 [local ] code_executor
2026-04-15 15:35:21 [INFO ] agent.MCP.SkillRegistry │ 🔌 SkillRegistry 已关闭所有连接
2026-04-15 15:35:23 [WARNING ] agent.TOOL.SSHDocker │ ⚠️ paramiko 未安装，请执行: pip install paramiko>=3.0.0
2026-04-15 15:35:23 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-autofix
2026-04-15 15:35:23 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-browser
2026-04-15 15:35:23 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-explorer
2026-04-15 15:35:23 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-oneshot
2026-04-15 15:35:23 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-usage
2026-04-15 15:35:23 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: smart-search
2026-04-15 15:35:23 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: calculator
2026-04-15 15:35:23 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: web_search
2026-04-15 15:35:23 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: file_reader
2026-04-15 15:35:23 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: code_executor
2026-04-15 15:35:23 [INFO ] agent.MCP.SkillRegistry │ 🌐 开始连接在线 MCP Skills，数量=1
2026-04-15 15:35:23 [INFO ] agent.MCP.SkillClient │ 🌐 连接在线 MCP Skill: [hexstrike-ai]
传输协议: stdio
地址 : python3
超时 : 300s
重试 : 2 次
2026-04-15 15:35:23 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 15:35:23 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=19588
2026-04-15 15:35:23 [WARNING ] agent.MCP.SkillClient │ ⚠️ 连接失败 (attempt 1/3)，1s 后重试: stdio 子进程无响应 skill=hexstrike-ai method=initialize
2026-04-15 15:35:24 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 15:35:24 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 15:35:24 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=19208
2026-04-15 15:35:24 [WARNING ] agent.MCP.SkillClient │ ⚠️ 连接失败 (attempt 2/3)，2s 后重试: stdio 子进程无响应 skill=hexstrike-ai method=initialize
2026-04-15 15:35:26 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 15:35:26 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 15:35:26 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=19680
2026-04-15 15:35:26 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 15:35:26 [ERROR ] agent.MCP.SkillRegistry │ ❌ Skill [hexstrike-ai] 连接失败，跳过
错误: ❌ MCP Skill [hexstrike-ai] 连接失败（已重试 2 次）
最后错误: stdio 子进程无响应 skill=hexstrike-ai method=initialize
2026-04-15 15:35:26 [DEBUG ] agent.MCP.SkillClient │ 🔌 MCP Skill [hexstrike-ai] 已断开
2026-04-15 15:35:26 [INFO ] agent.MCP.SkillRegistry │ 📊 SkillRegistry 初始化完成
本地工具 : 4 个 ['calculator', 'web_search', 'file_reader', 'code_executor']
远端工具 : 6 个 ['opencli-autofix', 'opencli-browser', 'opencli-explorer', 'opencli-oneshot', 'opencli-usage', 'smart-search']
2026-04-15 15:44:33 [WARNING ] agent.TOOL.SSHDocker │ ⚠️ paramiko 未安装，请执行: pip install paramiko>=3.0.0
2026-04-15 15:44:33 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-autofix
2026-04-15 15:44:33 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-autofix', 'description': 'Automatically fix broken OpenCLI adapters when commands fail. Load this skill when an opencli command fails — it guides you through diagnosing the failure via OPENCLI_DIAGNOSTIC, patching the adapter, retrying, and filing an upstream GitHub issue after a verified fix. Works with any AI agent.', 'allowed-tools': 'Bash(opencli:*), Bash(gh:*), Read, Edit, Write'}
2026-04-15 15:44:33 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-browser
2026-04-15 15:44:33 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-browser', 'description': 'Make websites accessible for AI agents. Navigate, click, type, extract, wait — using Chrome with existing login sessions. No LLM API key needed.', 'allowed-tools': 'Bash(opencli:*), Read, Edit, Write'}
2026-04-15 15:44:33 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-explorer
2026-04-15 15:44:33 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-explorer', 'description': 'Use when creating a new OpenCLI adapter from scratch, adding support for a new website or platform, exploring a site\'s API endpoints via browser DevTools, or when a user asks to automatically generate a CLI for a website (e.g. "帮我生成 xxx.com 的 cli"). Covers automated generation, API discovery workflow, authentication strategy selection, TS adapter writing, and testing.', 'tags': ['opencli', 'adapter', 'browser', 'api-discovery', 'cli', 'web-scraping', 'automation', 'generate']}
2026-04-15 15:44:33 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-oneshot
2026-04-15 15:44:33 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-oneshot', 'description': 'Use when quickly generating a single OpenCLI command from a specific URL and goal description. 4-step process — open page, capture API, write TS adapter, test. For full site exploration, use opencli-explorer instead.', 'tags': ['opencli', 'adapter', 'quick-start', 'ts', 'cli', 'one-shot', 'automation']}
2026-04-15 15:44:33 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-usage
2026-04-15 15:44:33 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-usage', 'description': 'Use when running OpenCLI commands to interact with websites (Bilibili, Twitter, Reddit, Xiaohongshu, etc.), desktop apps (Cursor, Notion), or public APIs (HackerNews, arXiv). Covers installation, command reference, and output formats for 87+ adapters.', 'version': '1.7.0', 'author': 'jackwener', 'tags': ['opencli', 'cli', 'browser', 'web', 'chrome-extension', 'cdp', 'bilibili', 'twitter', 'reddit', 'xiaohongshu', 'github', 'youtube', 'AI', 'agent', 'automation']}
2026-04-15 15:44:33 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: smart-search
2026-04-15 15:44:33 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'smart-search', 'description': '基于 opencli 命令的智能搜索路由器。当用户想要搜索、查询、查找或研究信息时，尤其是涉及指定网站、社交媒体、技术资料、新闻、购物、旅游、求职、金融或中文内容时，务必使用此 skill'}
2026-04-15 15:44:33 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: calculator
2026-04-15 15:44:33 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: web_search
2026-04-15 15:44:33 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: file_reader
2026-04-15 15:44:33 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: code_executor
2026-04-15 15:44:33 [INFO ] agent.MCP.SkillRegistry │ 🌐 开始连接在线 MCP Skills，数量=1
2026-04-15 15:44:33 [INFO ] agent.MCP.SkillClient │ 🌐 连接在线 MCP Skill: [hexstrike-ai]
传输协议: stdio
地址 : python3
超时 : 300s
重试 : 2 次
2026-04-15 15:44:33 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 15:44:33 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=19376
2026-04-15 15:44:33 [WARNING ] agent.MCP.SkillClient │ ⚠️ 连接失败 (attempt 1/3)，1s 后重试: stdio 子进程无响应 skill=hexstrike-ai method=initialize
2026-04-15 15:44:34 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 15:44:34 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 15:44:34 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=18048
2026-04-15 15:44:34 [WARNING ] agent.MCP.SkillClient │ ⚠️ 连接失败 (attempt 2/3)，2s 后重试: stdio 子进程无响应 skill=hexstrike-ai method=initialize
2026-04-15 15:44:36 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 15:44:36 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 15:44:36 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=19960
2026-04-15 15:44:36 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 15:44:36 [ERROR ] agent.MCP.SkillRegistry │ ❌ Skill [hexstrike-ai] 连接失败，跳过
错误: ❌ MCP Skill [hexstrike-ai] 连接失败（已重试 2 次）
最后错误: stdio 子进程无响应 skill=hexstrike-ai method=initialize
2026-04-15 15:44:36 [DEBUG ] agent.MCP.SkillClient │ 🔌 MCP Skill [hexstrike-ai] 已断开
2026-04-15 15:44:36 [INFO ] agent.MCP.SkillRegistry │ 📊 SkillRegistry 初始化完成
本地工具 : 4 个 ['calculator', 'web_search', 'file_reader', 'code_executor']
远端工具 : 0 个 []
2026-04-15 15:44:37 [INFO ] agent.Agent │ 🤖 Agent 初始化完成
LLM : openai / gpt-4o
工具总数 : 4 个
最大步数 : 10
工具列表 :
🔵 [local ] calculator
🔵 [local ] web_search
🔵 [local ] file_reader
🔵 [local ] code_executor
2026-04-15 16:09:20 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-autofix
2026-04-15 16:09:20 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-autofix', 'description': 'Automatically fix broken OpenCLI adapters when commands fail. Load this skill when an opencli command fails — it guides you through diagnosing the failure via OPENCLI_DIAGNOSTIC, patching the adapter, retrying, and filing an upstream GitHub issue after a verified fix. Works with any AI agent.', 'allowed-tools': 'Bash(opencli:*), Bash(gh:*), Read, Edit, Write'}
2026-04-15 16:09:20 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-browser
2026-04-15 16:09:20 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-browser', 'description': 'Make websites accessible for AI agents. Navigate, click, type, extract, wait — using Chrome with existing login sessions. No LLM API key needed.', 'allowed-tools': 'Bash(opencli:*), Read, Edit, Write'}
2026-04-15 16:09:20 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-explorer
2026-04-15 16:09:20 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-explorer', 'description': 'Use when creating a new OpenCLI adapter from scratch, adding support for a new website or platform, exploring a site\'s API endpoints via browser DevTools, or when a user asks to automatically generate a CLI for a website (e.g. "帮我生成 xxx.com 的 cli"). Covers automated generation, API discovery workflow, authentication strategy selection, TS adapter writing, and testing.', 'tags': ['opencli', 'adapter', 'browser', 'api-discovery', 'cli', 'web-scraping', 'automation', 'generate']}
2026-04-15 16:09:20 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-oneshot
2026-04-15 16:09:20 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-oneshot', 'description': 'Use when quickly generating a single OpenCLI command from a specific URL and goal description. 4-step process — open page, capture API, write TS adapter, test. For full site exploration, use opencli-explorer instead.', 'tags': ['opencli', 'adapter', 'quick-start', 'ts', 'cli', 'one-shot', 'automation']}
2026-04-15 16:09:20 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-usage
2026-04-15 16:09:20 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-usage', 'description': 'Use when running OpenCLI commands to interact with websites (Bilibili, Twitter, Reddit, Xiaohongshu, etc.), desktop apps (Cursor, Notion), or public APIs (HackerNews, arXiv). Covers installation, command reference, and output formats for 87+ adapters.', 'version': '1.7.0', 'author': 'jackwener', 'tags': ['opencli', 'cli', 'browser', 'web', 'chrome-extension', 'cdp', 'bilibili', 'twitter', 'reddit', 'xiaohongshu', 'github', 'youtube', 'AI', 'agent', 'automation']}
2026-04-15 16:09:20 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: smart-search
2026-04-15 16:09:20 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'smart-search', 'description': '基于 opencli 命令的智能搜索路由器。当用户想要搜索、查询、查找或研究信息时，尤其是涉及指定网站、社交媒体、技术资料、新闻、购物、旅游、求职、金融或中文内容时，务必使用此 skill'}
2026-04-15 16:09:28 [INFO ] agent.MCP.SkillRegistry │ 🔧 注册本地工具: ['calculator', 'web_search', 'file_reader', 'code_executor']
2026-04-15 16:10:46 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-autofix
2026-04-15 16:10:46 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-autofix', 'description': 'Automatically fix broken OpenCLI adapters when commands fail. Load this skill when an opencli command fails — it guides you through diagnosing the failure via OPENCLI_DIAGNOSTIC, patching the adapter, retrying, and filing an upstream GitHub issue after a verified fix. Works with any AI agent.', 'allowed-tools': 'Bash(opencli:*), Bash(gh:*), Read, Edit, Write'}
2026-04-15 16:10:46 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-browser
2026-04-15 16:10:46 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-browser', 'description': 'Make websites accessible for AI agents. Navigate, click, type, extract, wait — using Chrome with existing login sessions. No LLM API key needed.', 'allowed-tools': 'Bash(opencli:*), Read, Edit, Write'}
2026-04-15 16:10:46 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-explorer
2026-04-15 16:10:46 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-explorer', 'description': 'Use when creating a new OpenCLI adapter from scratch, adding support for a new website or platform, exploring a site\'s API endpoints via browser DevTools, or when a user asks to automatically generate a CLI for a website (e.g. "帮我生成 xxx.com 的 cli"). Covers automated generation, API discovery workflow, authentication strategy selection, TS adapter writing, and testing.', 'tags': ['opencli', 'adapter', 'browser', 'api-discovery', 'cli', 'web-scraping', 'automation', 'generate']}
2026-04-15 16:10:46 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-oneshot
2026-04-15 16:10:46 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-oneshot', 'description': 'Use when quickly generating a single OpenCLI command from a specific URL and goal description. 4-step process — open page, capture API, write TS adapter, test. For full site exploration, use opencli-explorer instead.', 'tags': ['opencli', 'adapter', 'quick-start', 'ts', 'cli', 'one-shot', 'automation']}
2026-04-15 16:10:46 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-usage
2026-04-15 16:10:46 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-usage', 'description': 'Use when running OpenCLI commands to interact with websites (Bilibili, Twitter, Reddit, Xiaohongshu, etc.), desktop apps (Cursor, Notion), or public APIs (HackerNews, arXiv). Covers installation, command reference, and output formats for 87+ adapters.', 'version': '1.7.0', 'author': 'jackwener', 'tags': ['opencli', 'cli', 'browser', 'web', 'chrome-extension', 'cdp', 'bilibili', 'twitter', 'reddit', 'xiaohongshu', 'github', 'youtube', 'AI', 'agent', 'automation']}
2026-04-15 16:10:46 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: smart-search
2026-04-15 16:10:46 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'smart-search', 'description': '基于 opencli 命令的智能搜索路由器。当用户想要搜索、查询、查找或研究信息时，尤其是涉及指定网站、社交媒体、技术资料、新闻、购物、旅游、求职、金融或中文内容时，务必使用此 skill'}
2026-04-15 16:10:46 [INFO ] agent.MCP.SkillRegistry │ 🔧 注册本地工具: ['calculator', 'web_search', 'file_reader', 'code_executor']
2026-04-15 16:10:53 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: calculator
2026-04-15 16:10:58 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: web_search
2026-04-15 16:10:58 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: file_reader
2026-04-15 16:10:58 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: code_executor
2026-04-15 16:12:30 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-autofix
2026-04-15 16:12:30 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-autofix', 'description': 'Automatically fix broken OpenCLI adapters when commands fail. Load this skill when an opencli command fails — it guides you through diagnosing the failure via OPENCLI_DIAGNOSTIC, patching the adapter, retrying, and filing an upstream GitHub issue after a verified fix. Works with any AI agent.', 'allowed-tools': 'Bash(opencli:*), Bash(gh:*), Read, Edit, Write'}
2026-04-15 16:12:30 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-browser
2026-04-15 16:12:30 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-browser', 'description': 'Make websites accessible for AI agents. Navigate, click, type, extract, wait — using Chrome with existing login sessions. No LLM API key needed.', 'allowed-tools': 'Bash(opencli:*), Read, Edit, Write'}
2026-04-15 16:12:30 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-explorer
2026-04-15 16:12:30 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-explorer', 'description': 'Use when creating a new OpenCLI adapter from scratch, adding support for a new website or platform, exploring a site\'s API endpoints via browser DevTools, or when a user asks to automatically generate a CLI for a website (e.g. "帮我生成 xxx.com 的 cli"). Covers automated generation, API discovery workflow, authentication strategy selection, TS adapter writing, and testing.', 'tags': ['opencli', 'adapter', 'browser', 'api-discovery', 'cli', 'web-scraping', 'automation', 'generate']}
2026-04-15 16:12:30 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-oneshot
2026-04-15 16:12:30 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-oneshot', 'description': 'Use when quickly generating a single OpenCLI command from a specific URL and goal description. 4-step process — open page, capture API, write TS adapter, test. For full site exploration, use opencli-explorer instead.', 'tags': ['opencli', 'adapter', 'quick-start', 'ts', 'cli', 'one-shot', 'automation']}
2026-04-15 16:12:30 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: opencli-usage
2026-04-15 16:12:30 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'opencli-usage', 'description': 'Use when running OpenCLI commands to interact with websites (Bilibili, Twitter, Reddit, Xiaohongshu, etc.), desktop apps (Cursor, Notion), or public APIs (HackerNews, arXiv). Covers installation, command reference, and output formats for 87+ adapters.', 'version': '1.7.0', 'author': 'jackwener', 'tags': ['opencli', 'cli', 'browser', 'web', 'chrome-extension', 'cdp', 'bilibili', 'twitter', 'reddit', 'xiaohongshu', 'github', 'youtube', 'AI', 'agent', 'automation']}
2026-04-15 16:12:30 [INFO ] agent.MCP.SkillRegistry │ 📦 加载技能: smart-search
2026-04-15 16:12:30 [WARNING ] agent.MCP.SkillRegistry │ ⚠️ 工具实例缺少 name 属性，跳过: {'name': 'smart-search', 'description': '基于 opencli 命令的智能搜索路由器。当用户想要搜索、查询、查找或研究信息时，尤其是涉及指定网站、社交媒体、技术资料、新闻、购物、旅游、求职、金融或中文内容时，务必使用此 skill'}
2026-04-15 16:12:30 [INFO ] agent.MCP.SkillRegistry │ 🔧 注册本地工具: ['calculator', 'web_search', 'file_reader', 'code_executor']
2026-04-15 16:12:30 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: calculator
2026-04-15 16:12:30 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: web_search
2026-04-15 16:12:30 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: file_reader
2026-04-15 16:12:30 [DEBUG ] agent.MCP.SkillRegistry │ 📌 注册本地工具: code_executor
2026-04-15 16:12:30 [INFO ] agent.MCP.SkillRegistry │ 🌐 开始连接在线 MCP Skills，数量=1
2026-04-15 16:12:30 [INFO ] agent.MCP.SkillClient │ 🌐 连接在线 MCP Skill: [hexstrike-ai]
传输协议: stdio
地址 : python3
超时 : 300s
重试 : 2 次
2026-04-15 16:12:30 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 16:12:33 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=18196
2026-04-15 16:12:37 [WARNING ] agent.MCP.SkillClient │ ⚠️ 连接失败 (attempt 1/3)，1s 后重试: [Errno 22] Invalid argument
2026-04-15 16:12:38 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 16:12:38 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 16:12:38 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=14612
2026-04-15 16:12:38 [WARNING ] agent.MCP.SkillClient │ ⚠️ 连接失败 (attempt 2/3)，2s 后重试: stdio 子进程无响应 skill=hexstrike-ai method=initialize
2026-04-15 16:12:40 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 16:12:40 [INFO ] agent.MCP.SkillClient │ 🔌 stdio 启动子进程: python3 /Users/sontolau/Applications/hexstrike-ai/hexstrike_mcp.py --server http://localhost:8999
2026-04-15 16:12:40 [INFO ] agent.MCP.SkillClient │ ✅ stdio 子进程已启动 PID=7028
2026-04-15 16:12:40 [DEBUG ] agent.MCP.SkillClient │ 🔌 stdio 子进程已关闭 skill=hexstrike-ai
2026-04-15 16:12:40 [ERROR ] agent.MCP.SkillRegistry │ ❌ Skill [hexstrike-ai] 连接失败，跳过
错误: ❌ MCP Skill [hexstrike-ai] 连接失败（已重试 2 次）
最后错误: stdio 子进程无响应 skill=hexstrike-ai method=initialize
2026-04-15 16:12:40 [DEBUG ] agent.MCP.SkillClient │ 🔌 MCP Skill [hexstrike-ai] 已断开
2026-04-15 16:12:40 [INFO ] agent.MCP.SkillRegistry │ 📊 SkillRegistry 初始化完成
本地工具 : 4 个 ['calculator', 'web_search', 'file_reader', 'code_executor']
远端工具 : 0 个 []
2026-04-15 16:12:42 [INFO ] agent.Agent │ 🤖 Agent 初始化完成
LLM : openai / gpt-4o
工具总数 : 4 个
最大步数 : 10
工具列表 :
🔵 [local ] calculator
🔵 [local ] web_search
🔵 [local ] file_reader
🔵 [local ] code_executor
2026-04-15 16:19:35 [INFO ] agent.MCP.SkillRegistry │ 🔌 SkillRegistry 已关闭所有连接

148
main.py
View File

@ -60,78 +60,78 @@ def run_mcp_server() -> None:
server.run_stdio() server.run_stdio()
def run_check() -> None: # def run_check() -> None:
"""检查配置和依赖完整性""" # """检查配置和依赖完整性"""
print("=" * 60) # print("=" * 60)
print(" 🔍 项目依赖检查") # print(" 🔍 项目依赖检查")
print("=" * 60) # print("=" * 60)
#
checks = [ # checks = [
("pyyaml", "yaml", "pip install pyyaml"), # ("pyyaml", "yaml", "pip install pyyaml"),
("openai", "openai", "pip install openai>=1.0.0"), # ("openai", "openai", "pip install openai>=1.0.0"),
("httpx", "httpx", "pip install httpx>=0.27.0"), # ("httpx", "httpx", "pip install httpx>=0.27.0"),
("httpx-sse", "httpx_sse", "pip install httpx-sse>=0.4.0"), # ("httpx-sse", "httpx_sse", "pip install httpx-sse>=0.4.0"),
("paramiko", "paramiko", "pip install paramiko>=3.0.0"), # ("paramiko", "paramiko", "pip install paramiko>=3.0.0"),
] # ]
#
all_ok = True # all_ok = True
for pkg_name, import_name, install_cmd in checks: # for pkg_name, import_name, install_cmd in checks:
try: # try:
__import__(import_name) # __import__(import_name)
print(f" ✅ {pkg_name:<15} 已安装") # print(f" ✅ {pkg_name:<15} 已安装")
except ImportError: # except ImportError:
print(f" ❌ {pkg_name:<15} 未安装 → {install_cmd}") # print(f" ❌ {pkg_name:<15} 未安装 → {install_cmd}")
all_ok = False # all_ok = False
#
print() # print()
#
# 配置检查 # # 配置检查
try: # try:
from config.settings import settings # from config.settings import settings
print(" ✅ config/settings.py 加载成功") # print(" ✅ config/settings.py 加载成功")
print(f" LLM : {settings.llm.provider} / {settings.llm.model_name}") # print(f" LLM : {settings.llm.provider} / {settings.llm.model_name}")
print(f" 本地工具: {settings.mcp.enabled_tools}") # print(f" 本地工具: {settings.mcp.enabled_tools}")
skills = settings.enabled_mcp_skills # skills = settings.enabled_mcp_skills
if skills: # if skills:
print(f" 在线Skill: {[s.name for s in skills]}") # print(f" 在线Skill: {[s.name for s in skills]}")
else: # else:
print(" 在线Skill: (未配置)") # print(" 在线Skill: (未配置)")
except Exception as e: # except Exception as e:
print(f" ❌ 配置加载失败: {e}") # print(f" ❌ 配置加载失败: {e}")
all_ok = False # all_ok = False
#
print() # print()
#
# 工具注册检查 # # 工具注册检查
try: # try:
from mcp.skill_registry import SkillRegistry # from mcp.skill_registry import SkillRegistry
from tools.calculator import CalculatorTool # from tools.calculator import CalculatorTool
from tools.code_executor import CodeExecutorTool # from tools.code_executor import CodeExecutorTool
from tools.file_reader import FileReaderTool # from tools.file_reader import FileReaderTool
from tools.ssh_docker import SSHDockerTool # from tools.ssh_docker import SSHDockerTool
from tools.static_analyzer import StaticAnalyzerTool # from tools.static_analyzer import StaticAnalyzerTool
from tools.web_search import WebSearchTool # from tools.web_search import WebSearchTool
#
registry = SkillRegistry() # registry = SkillRegistry()
registry.register_local_many( # registry.register_local_many(
CalculatorTool(), WebSearchTool(), FileReaderTool(), # CalculatorTool(), WebSearchTool(), FileReaderTool(),
CodeExecutorTool(), StaticAnalyzerTool(), SSHDockerTool(), # CodeExecutorTool(), StaticAnalyzerTool(), SSHDockerTool(),
) # )
tools = registry.list_all_tools() # tools = registry.list_all_tools()
print(f" ✅ 本地工具注册 共 {len(tools)} 个:") # print(f" ✅ 本地工具注册 共 {len(tools)} 个:")
for t in tools: # for t in tools:
print(f" 🔵 {t['name']}: {t['description'][:50]}") # print(f" 🔵 {t['name']}: {t['description'][:50]}")
except Exception as e: # except Exception as e:
print(f" ❌ 工具注册失败: {e}") # print(f" ❌ 工具注册失败: {e}")
all_ok = False # all_ok = False
#
print() # print()
print("=" * 60) # print("=" * 60)
if all_ok: # if all_ok:
print(" ✅ 所有检查通过，项目可正常运行") # print(" ✅ 所有检查通过，项目可正常运行")
else: # else:
print(" ⚠️ 存在问题，请按提示安装缺失依赖") # print(" ⚠️ 存在问题，请按提示安装缺失依赖")
print("=" * 60) # print("=" * 60)
def main() -> None: def main() -> None:
@ -151,8 +151,8 @@ def main() -> None:
run_agent() run_agent()
case "mcp": case "mcp":
run_mcp_server() run_mcp_server()
case "check": # case "check":
run_check() # run_check()
if __name__ == "__main__": if __name__ == "__main__":

58
mcp/mcp_server.py
View File

@ -6,30 +6,12 @@ mcp/mcp_server.py
import json import json
import sys import sys
from typing import Any from typing import Any
from config.settings import settings from config.settings import settings
from mcp.skill_registry import SkillRegistry from mcp.skill_registry import SkillRegistry
from tools.calculator import CalculatorTool
from tools.code_executor import CodeExecutorTool
from tools.file_reader import FileReaderTool
from tools.ssh_docker import SSHDockerTool
from tools.static_analyzer import StaticAnalyzerTool
from tools.web_search import WebSearchTool
from utils.logger import get_logger from utils.logger import get_logger
logger = get_logger("MCP.Server") logger = get_logger("MCP.Server")
# 本地工具类映射表
_LOCAL_TOOL_CLASSES: dict[str, type] = {
"calculator": CalculatorTool,
"web_search": WebSearchTool,
"file_reader": FileReaderTool,
"code_executor": CodeExecutorTool,
"static_analyzer": StaticAnalyzerTool,
"ssh_docker": SSHDockerTool,
}
class MCPServer: class MCPServer:
""" """
本地 MCP Server 本地 MCP Server
@ -48,17 +30,37 @@ class MCPServer:
def _setup(self) -> None: def _setup(self) -> None:
"""初始化：注册本地工具 + 连接在线 Skill""" """初始化：注册本地工具 + 连接在线 Skill"""
# ── 注册本地工具 ────────────────────────────────────── # ── 注册本地工具 ──────────────────────────────────────
enabled = settings.mcp.enabled_tools # enabled = settings.mcp.enabled_tools
logger.info(f"🔧 注册本地工具: {enabled}") # logger.info(f"🔧 注册本地工具: {enabled}")
for tool_name in enabled: # for tool_name in enabled:
cls = _LOCAL_TOOL_CLASSES.get(tool_name) # tool_path = f"tools/{tool_name}.py"
if cls: # if not os.path.exists(tool_path):
self.registry.register_local(cls()) # continue
else: # # module_name = filename[:-3] # 去掉 .py 后缀
logger.warning(f"⚠️ 未知工具: {tool_name}，跳过") # # module_path = os.path.join(directory_path, filename)
#
# # 动态加载模块
# spec = importlib.util.spec_from_file_location(tool_name, tool_path)
# module = importlib.util.module_from_spec(spec)
# spec.loader.exec_module(module)
#
# if not hasattr(module, "Tool"):
# continue
#
# cls = getattr(module, 'Tool')
# if not isinstance(cls, BaseTool):
# continue
#
# # tool_classes.append(tool_class)
# # cls = _LOCAL_TOOL_CLASSES.get(tool_name)
# if cls:
# self.registry.register_local(cls())
# else:
# logger.warning(f"⚠️ 未知工具: {tool_name}，跳过")
self.registry.load_local_tools()
self.registry.load_local_skills(settings.skills_directory)
# ── 连接在线 MCP Skill ──────────────────────────────── # ── 连接在线 MCP Skill ────────────────────────────────
skill_map = self.registry.connect_skills() skill_map = self.registry.connect_mcp_skills()
if skill_map: if skill_map:
logger.info( logger.info(
"🌐 在线 Skill 注册汇总:\n" + "🌐 在线 Skill 注册汇总:\n" +

3
mcp/skill_loader.py
View File

@ -14,11 +14,10 @@ class SkillLoader:
def load_skills_from_directory(directory: str) -> Dict[str, Any]: def load_skills_from_directory(directory: str) -> Dict[str, Any]:
skills = {} skills = {}
for skill_directory in os.listdir(directory): for skill_directory in os.listdir(directory):
skill_md = f"{skill_directory}/SKILL.md" skill_md = f"{directory}/{skill_directory}/SKILL.md"
if not os.path.isfile(skill_md): if not os.path.isfile(skill_md):
logger.warning(f"未在{skill_directory}发现SKILL.md文件") logger.warning(f"未在{skill_directory}发现SKILL.md文件")
continue continue
skill_name = os.path.basename(skill_directory) # 去掉 .md 后缀
skill_info = SkillLoader.load_skill(skill_md) skill_info = SkillLoader.load_skill(skill_md)
if not skill_info: if not skill_info:
continue continue

36
mcp/skill_registry.py
View File

@ -8,7 +8,8 @@ mcp/skill_registry.py
- dispatch() → 根据工具名路由到本地或远端执行 - dispatch() → 根据工具名路由到本地或远端执行
- refresh_skills() → 重新拉取在线 Skill 工具列表 - refresh_skills() → 重新拉取在线 Skill 工具列表
""" """
import importlib
import os
import time import time
from dataclasses import dataclass from dataclasses import dataclass
from typing import Any from typing import Any
@ -16,6 +17,7 @@ from typing import Any
from config.settings import settings from config.settings import settings
from mcp.mcp_skill_client import MCPSkillClient, RemoteTool, ToolCallResult from mcp.mcp_skill_client import MCPSkillClient, RemoteTool, ToolCallResult
from mcp.skill_loader import SkillLoader from mcp.skill_loader import SkillLoader
from tools.base_tool import BaseTool
from utils.logger import get_logger from utils.logger import get_logger
logger = get_logger("MCP.SkillRegistry") logger = get_logger("MCP.SkillRegistry")
@ -82,15 +84,37 @@ class SkillRegistry:
self._remote: dict[str, tuple[MCPSkillClient, RemoteTool]] = {} self._remote: dict[str, tuple[MCPSkillClient, RemoteTool]] = {}
# 在线 Skill 客户端列表（用于生命周期管理） # 在线 Skill 客户端列表（用于生命周期管理）
self._clients: list[MCPSkillClient] = [] self._clients: list[MCPSkillClient] = []
# 指定 SKILL.md 文件所在目录
self.load_skills_from_md(settings.skills_directory)
def load_skills_from_md(self, directory: str) -> None:
def load_local_skills(self, directory: str) -> None:
skills = SkillLoader.load_skills_from_directory(directory) skills = SkillLoader.load_skills_from_directory(directory)
for skill_name, skill_info in skills.items(): for skill_name, skill_info in skills.items():
logger.info(f"📦 加载技能: {skill_name}") logger.info(f"📦 加载技能: {skill_name}")
# 将技能注册到远端工具表 # 将技能注册到远端工具表
self._remote[skill_name] = skill_info #self._remote[skill_name] = skill_info
self.register_local(skill_info)
def load_local_tools(self):
enabled = settings.mcp.enabled_tools
logger.info(f"🔧 注册本地工具: {enabled}")
for tool_name in enabled:
tool_path = f"tools/{tool_name}.py"
if not os.path.exists(tool_path):
continue
# 动态加载模块
spec = importlib.util.spec_from_file_location(tool_name, tool_path)
module = importlib.util.module_from_spec(spec)
spec.loader.exec_module(module)
if not hasattr(module, "Tool"):
continue
cls = getattr(module, 'Tool')
if not issubclass(cls, BaseTool):
continue
if cls:
self.register_local(cls())
else:
logger.warning(f"⚠️ 未知工具: {tool_name}，跳过")
# ── 注册本地工具 ────────────────────────────────────────── # ── 注册本地工具 ──────────────────────────────────────────
@ -117,7 +141,7 @@ class SkillRegistry:
# ── 连接在线 MCP Skill ──────────────────────────────────── # ── 连接在线 MCP Skill ────────────────────────────────────
def connect_skills(self) -> dict[str, list[str]]: def connect_mcp_skills(self) -> dict[str, list[str]]:
""" """
连接所有 config.yaml 中 enabled=true 的在线 MCP Skill 连接所有 config.yaml 中 enabled=true 的在线 MCP Skill
并将其工具注册到远端工具表 并将其工具注册到远端工具表

3
tools/calculator.py
View File

@ -9,6 +9,7 @@ import operator
from typing import Any from typing import Any
from config.settings import settings from config.settings import settings
from tools.base_tool import BaseTool
from utils.logger import get_logger from utils.logger import get_logger
logger = get_logger("TOOL.Calculator") logger = get_logger("TOOL.Calculator")
@ -18,7 +19,7 @@ def _cfg(key: str, fallback=None):
return settings.tools['calculator'].get(key, fallback) return settings.tools['calculator'].get(key, fallback)
class CalculatorTool: class Tool(BaseTool):
name = "calculator" name = "calculator"
description = ( description = (
"执行数学计算，支持四则运算、幂运算、开方、三角函数、对数等。" "执行数学计算，支持四则运算、幂运算、开方、三角函数、对数等。"

3
tools/code_executor.py
View File

@ -12,6 +12,7 @@ import traceback
from contextlib import redirect_stderr, redirect_stdout from contextlib import redirect_stderr, redirect_stdout
from config.settings import settings from config.settings import settings
from tools.base_tool import BaseTool
from utils.logger import get_logger from utils.logger import get_logger
logger = get_logger("TOOL.CodeExecutor") logger = get_logger("TOOL.CodeExecutor")
@ -21,7 +22,7 @@ def _cfg(key: str, fallback=None):
return settings.tools['code_executor'].get(key, fallback) return settings.tools['code_executor'].get(key, fallback)
class CodeExecutorTool: class Tool(BaseTool):
name = "code_executor" name = "code_executor"
description = ( description = (
"在安全沙箱中执行 Python 代码片段，返回标准输出和执行结果。" "在安全沙箱中执行 Python 代码片段，返回标准输出和执行结果。"

3
tools/file_reader.py
View File

@ -10,6 +10,7 @@ import json
from pathlib import Path from pathlib import Path
from config.settings import settings from config.settings import settings
from tools.base_tool import BaseTool
from utils.logger import get_logger from utils.logger import get_logger
logger = get_logger("TOOL.FileReader") logger = get_logger("TOOL.FileReader")
@ -19,7 +20,7 @@ def _cfg(key: str, fallback=None):
return settings.tools['file_reader'].get(key, fallback) return settings.tools['file_reader'].get(key, fallback)
class FileReaderTool: class Tool(BaseTool):
name = "file_reader" name = "file_reader"
description = ( description = (
"读取本地文件内容，支持 .txt / .md / .py / .json / .csv / .yaml / .log 等文本文件。" "读取本地文件内容，支持 .txt / .md / .py / .json / .csv / .yaml / .log 等文本文件。"

3
tools/ssh_docker.py
View File

@ -10,6 +10,7 @@ import time
from dataclasses import dataclass, field from dataclasses import dataclass, field
from config.settings import settings from config.settings import settings
from tools.base_tool import BaseTool
from utils.logger import get_logger from utils.logger import get_logger
logger = get_logger("TOOL.SSHDocker") logger = get_logger("TOOL.SSHDocker")
@ -324,7 +325,7 @@ class DockerExecutor:
# 主工具类 # 主工具类
# ════════════════════════════════════════════════════════════════ # ════════════════════════════════════════════════════════════════
class SSHDockerTool: class Tool(BaseTool):
""" """
SSH 远程 Docker 部署工具 SSH 远程 Docker 部署工具
所有配置均通过 settings.tools['ssh_docker'][key] 读取 所有配置均通过 settings.tools['ssh_docker'][key] 读取

3
tools/static_analyzer.py
View File

@ -13,6 +13,7 @@ from dataclasses import dataclass, field
from pathlib import Path from pathlib import Path
from config.settings import settings from config.settings import settings
from tools.base_tool import BaseTool
from utils.logger import get_logger from utils.logger import get_logger
logger = get_logger("TOOL.StaticAnalyzer") logger = get_logger("TOOL.StaticAnalyzer")
@ -281,7 +282,7 @@ _TOOL_REGISTRY: dict[str, type] = {
# 主工具类 # 主工具类
# ════════════════════════════════════════════════════════════════ # ════════════════════════════════════════════════════════════════
class StaticAnalyzerTool: class Tool(BaseTool):
""" """
C/C++ 静态分析工具 C/C++ 静态分析工具
所有配置均通过 settings.tools['static_analyzer'][key] 读取 所有配置均通过 settings.tools['static_analyzer'][key] 读取

28
tools/tool_generator.py
View File

@ -1,28 +0,0 @@
from tools.base_tool import BaseTool, ToolResult
class ToolGeneratorTool(BaseTool):
name = "tool_generator"
description = "生成agent工具代码"
parameters = {
"name": {
"type": "string",
"description": "tool name",
},
"description": {
"type": "string",
"description": "tool description"
},
"parameters": {
"type": "object",
"description": """tool parameters descriptions"""
},
"code": {
"type": "string",
"description": "the code that the LLM generated"
}
}
def execute(self, name: str, description: str, parameters: dict, code: str) -> ToolResult:
pass

3
tools/web_search.py
View File

@ -9,6 +9,7 @@ import time
from dataclasses import dataclass, field from dataclasses import dataclass, field
from config.settings import settings from config.settings import settings
from tools.base_tool import BaseTool
from utils.logger import get_logger from utils.logger import get_logger
logger = get_logger("TOOL.WebSearch") logger = get_logger("TOOL.WebSearch")
@ -29,7 +30,7 @@ class SearchResult:
return f"[{self.rank}] {self.title}\n {self.url}\n {self.snippet}" return f"[{self.rank}] {self.title}\n {self.url}\n {self.snippet}"
class WebSearchTool: class Tool(BaseTool):
name = "web_search" name = "web_search"
description = ( description = (
"在互联网上搜索信息，返回相关网页的标题、链接和摘要。" "在互联网上搜索信息，返回相关网页的标题、链接和摘要。"