dsh-alpha-desk
by JingHao-Leon
Alpha Desk — a deepseek-harness (dsh) skill pack that turns an agent session into a compliance-first AI investment desk: multi-strategy fund backtesting via ai-hedge-fund, a tools/pre-execute risk gate, cron monitoring and thesis memory. 把 dsh 会话变成合规、可复现、可追责的 AI 投研工作台。
54stars
0forks
Python
Added 9/8/2026
View on GitHubDownload ZIPScan for vulnerabilities
Unlocks once the catalog security scan passes (runs nightly).
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
AI Agentsagent-skillsai-hedge-fundbacktestingdeepseekdeepseek-harnessdshfintechinvestment-researchllm-agentsmcp
Installation
# Add to your Claude Code skills
git clone https://github.com/JingHao-Leon/dsh-alpha-deskGetting Started
Guides for using ai agents skills like dsh-alpha-desk.
- Caveman: Cut Claude Token Use by 65%How agent-side prompt compression works, when to use it, and when not to.
- What is an AI Skills Marketplace?Definitions, how marketplaces work, and how to choose between them in 2026.
- Getting Started with AI SkillsFirst-time install walkthrough for Claude Code, Codex CLI, and ChatGPT.
Alpha Desk — deepseek-harness 的 AI 投研工作台
English | 中文
把 deepseek-harness（dsh）会话变成一个投研交易台： 多策略 AI 基金回测 + A股/港股技术分析 + 风控合规门禁 + 定时盯盘 + 投资记忆复盘。 本仓库是一个可直接安装的 dsh skill 包，也是 dsh 微内核扩展点（skill / hook / cron / memory）的完整示范。
免责声明：本项目仅供学习研究，不构成投资建议，暂不执行真实交易。
它解决什么问题
通用 agent 框架能"聊天"，但做投研缺四样东西：
- 可复现的引擎 —— 结论不能是模型即兴编的，必须来自真实数据与真实计算
- 合规边界 —— agent 权限越来越大，"永远碰不到真钱"要是架构保证，不是提示词约定
- 持续性 —— 投资观点要被记录、被定时检验、被事后追责
- 多市场 —— 美股、A股、港股各一套数据与逻辑
Alpha Desk 用 dsh 的四个扩展点各解决一个问题：
| 问题 | dsh 扩展点 | 本仓库实现 |
|---|---|---|
| 可复现引擎 | skill（section + inject()） |
skill/SKILL.md：编排 virattt/ai-hedge-fund CLI，stdout 纯 JSON，全部落盘 |
| 合规边界 | hook（tools/pre-execute waterfall） |
plugins/risk-gate：实盘下单/券商 API/凭证访问在分发前被单调拒绝 |
| 持续性 | cron + memory | SKILL.md 工作流四/五：盘前扫描、周末复盘、假设台账 |
| 多市场 | skill 组合 | 美股走 aihf；A股/港股联动 stock-technical-indicators 技能 |
| 权威数据 | agent 工具链 | plugins/ifind：iFinD 数据源（财报/公告/股东/预测/选股），经 Kimi agent-gw，免 iFinD 账号，凭证自解析 |
架构
用户（自然语言）
│
▼
deepseek-harness agent ── inject ──► skill/SKILL.md（本仓库）
│ │
│ tools/pre-execute ▼
│ ◄── risk-gate 插件（拒绝实盘） aihf CLI（ai-hedge-fund 引擎）
│ │ stdout: CycleRecord / BacktestResult JSON
│ ▼
│ records/*.json（落盘，供复盘）
│
├── cron：盘前/周末定时触发投研工作流
└── memory：投资假设台账，到期对照 records 复盘
底层引擎 ai-hedge-fund（MIT）把"基金"做成声明式 mandate：策略 pod、投资大师 alpha 模型（格雷厄姆/巴菲特/芒格/林奇/德鲁肯米勒 + 量化 PEAD）、风控限额、再平衡频率都是 YAML 数据；标的在运行时通过 --tickers 传入。aihf 原生支持 DeepSeek 作为推理 LLM——DeepSeek 模型跑在 DeepSeek harness 里驱动 AI 基金，全栈同构。
量化终端（terminal/）
把上面的 agent 能力装进一个同花顺风格的 Web 终端——A股自选实时报价、K线、右栏明细，中间是对话区,所有回复都过 risk-gate:
数据层用 vnpy 的 BarData/TickData 对象模型与 Gateway 语义封装腾讯免费行情(分钟级延迟),未来换 CTP/SimNow 是 drop-in 替换;agent 走 dsh --profile headless + risk-gate patch。只读研究终端,没有下单路径。 安装与 API 详见 terminal/README.md。
快速开始
# 1. 安装引擎（pipx 隔离安装）
pipx install aihf
# 2. 配置 key（二选一：shell 环境变量，或写入 ~/.hedge-fund/.env）
export FINANCIAL_DATASETS_API_KEY=... # 行情与基本面，financialdatasets.ai 有免费档
export DEEPSEEK_API_KEY=... # alpha 模型的推理 LLM
# 3. 手动验证引擎
aihf mandates/deep-value-weekly.yaml --tickers AAPL,MSFT,NVDA --backtest
# 4. 把 skill 装进 dsh（按你的 dsh 安装方式，将 skill/ 目录加入技能搜索路径，
# 例如软链到 ~/.agents/skills/alpha-desk）
ln -s "$PWD/skill" ~/.agents/skills/alpha-desk
# 5. （可选但推荐）启用 risk-gate 插件，见 plugins/risk-gate/
然后在 dsh 里直接说人话：
- 「用深度价值基金回测一下苹果、微软、英伟达过去一年」
- 「巴菲特和格雷厄姆对 NVDA 的看法有什么分歧？」
- 「每个交易日盘前帮我跑一遍持仓信号」
- 「上周你建议关注的几只票，现在复盘一下对错」
仓库结构
dsh-alpha-desk/
├── skill/SKILL.md # dsh 技能本体（触发条件、工作流、合规红线）
├── mandates/ # 三份开箱即用的基金委托书
│ ├── deep-value-weekly.yaml # 深度价值（格雷厄姆×2+巴菲特+芒格）周频
│ ├── fundamental-ls-market-neutral.yaml # 五大师多空市场中性，月频
│ └── inflections-daily.yaml # 宏观拐点（德鲁肯米勒+林奇）日频
├── plugins/risk-gate/ # dsh 风控钩子插件（tools/pre-execute）
├── plugins/ifind/ # iFinD 数据源（经 Kimi agent-gw，免 iFinD 账号）
├── terminal/ # 同花顺风格量化终端（vnpy 数据层 + FastAPI + React）
│ ├── backend/app/gateway_gtimg.py # 腾讯行情 → vnpy BarData/TickData
│ ├── backend/app/fundamentals.py # iFinD 基本面（agent-gw 直调 + 7 天缓存）
│ ├── backend/app/agent_bridge.py # dsh headless 桥（risk-gate 已挂载）
│ └── web/ # React + klinecharts 终端 UI
├── records/ # 运行记录落盘目录（git 忽略）
└── LICENSE # MIT
与主流 agent 框架的对比
| 能力 | Alpha Desk (dsh) | 裸 LangChain/CrewAI 编排 | 通用编码 agent + 提示词 |
|---|---|---|---|
| 合规边界 | 架构级：hook 在工具分发前单调拒绝 | 需自己包一层执行器 | 仅靠提示词，模型可绕过 |
| 投研引擎 | 成熟开源基金引擎，JSON 可复现 | 自行拼装 | 模型即兴生成，不可复现 |
| 定时盯盘 | dsh cron 扩展点，声明式 | 需外部调度器 | 无 |
| 观点追责 | memory + 落盘记录，到期自动复盘 | 自行实现状态层 | 无 |
| 热重载/生态 | dsh 插件 HMR，MCP/skill 生态复用 | — | — |
致谢与边界
- 引擎：virattt/ai-hedge-fund（MIT），本项目只做编排层，不 fork 其代码；其数据源 Financial Datasets 主要覆盖美股
- 运行时：deepseek-ai/deepseek-harness
- 数据源覆盖不了 A股/港股；回测结果不代表未来收益；本项目暂不执行真实交易
Frequently Asked Questions
What is dsh-alpha-desk?
dsh-alpha-desk is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by JingHao-Leon. Alpha Desk — a deepseek-harness (dsh) skill pack that turns an agent session into a compliance-first AI investment desk: multi-strategy fund backtesting via ai-hedge-fund, a tools/pre-execute risk gate, cron monitoring and thesis memory. 把 dsh 会话变成合规、可复现、可追责的 AI 投研工作台。. It has 54 GitHub stars.
Is dsh-alpha-desk safe to use?
dsh-alpha-desk's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
How do I install dsh-alpha-desk?
Clone the repository with "git clone https://github.com/JingHao-Leon/dsh-alpha-desk" and add it to your Claude Code skills directory (see the Installation section above).
What programming language is dsh-alpha-desk written in?
dsh-alpha-desk is primarily written in Python. It is open-source under JingHao-Leon on GitHub, so you can review or fork the full source.
Are there alternatives to dsh-alpha-desk?
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh dsh-alpha-desk against similar tools.
Comments (0)
to leave a comment.
No comments yet. Be the first to share your thoughts!
Related Skills
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
253,286
37,987
JavaScript
AI Agentsai-agentsanthropic
Skills for Real Engineers. Straight from my .agents directory.
250,518
21,171
Shell
AI Agents
An agentic skills framework & software development methodology that works.
234,966
20,863
Shell
AI Agentsaibrainstorming
Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations.
203,698
60,611
TypeScript
MCP Serversaiapis
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
185,940
28,768
JavaScript
AI Agentsai-agentsanthropic