量化AI和交易AI技能 是 AI Skill Hub 本期精选Agent工作流之一。综合评分 7.5 分,整体质量较高。我们推荐使用将其纳入你的 AI 工具库,帮助提升工作效率。
量化AI和交易AI技能 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
量化AI和交易AI技能 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一:pip 安装(推荐)
pip install abel-strategy-research-skills
# 方式二:虚拟环境安装(推荐生产环境)
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install abel-strategy-research-skills
# 方式三:从源码安装(获取最新功能)
git clone https://github.com/Abel-ai-lab/abel-strategy-research-skills
cd abel-strategy-research-skills
pip install -e .
# 验证安装
python -c "import abel_strategy_research_skills; print('安装成功')"
# 命令行使用
abel-strategy-research-skills --help
# 基本用法
abel-strategy-research-skills input_file -o output_file
# Python 代码中调用
import abel_strategy_research_skills
# 示例
result = abel_strategy_research_skills.process("input")
print(result)
# abel-strategy-research-skills 配置文件示例(config.yml) app: name: "abel-strategy-research-skills" debug: false log_level: "INFO" # 运行时指定配置文件 abel-strategy-research-skills --config config.yml # 或通过环境变量配置 export ABEL_STRATEGY_RESEARCH_SKILLS_API_KEY="your-key" export ABEL_STRATEGY_RESEARCH_SKILLS_OUTPUT_DIR="./output"
<p align="center"> <strong>Ask market questions, predict what moves next, and turn the strongest ideas into backtested strategy research with Predict Anything by Abel.</strong> </p>
<p align="center"> <a href="#codex">OpenAI Codex</a> | <a href="#claude-code">Claude Code</a> | <a href="#opencode">OpenCode</a> | <a href="#clawhub--openclaw">ClawHub</a> </p>
Predict Anything turns coding agents into market research agents. Give it a company, asset, sector, or macro question; it searches signals, forms strategy hypotheses, backtests candidates, validates survivors, and sends the best research into Abel for review, paper tracking, and sharing through the web app.
- From question to strategy Start with a plain market question. Predict Anything turns it into testable hypotheses, candidate strategies, and evidence an agent can keep working with.
- Prediction with evidence Look past surface correlations. Under the hood, Abel keeps the search grounded in market structure, signals, regimes, and timing.
- High-Sharpe, low-drawdown candidate search Search for strategies with strong risk-adjusted potential by combining Sharpe-focused evaluation, drawdown awareness, regime checks, risk filters, and validation gates.
- Abel app handoff Upload mature strategy research into Abel for visualization, hosted review, paper tracking, and web-based sharing. The skill supports research and validation; it does not place trades or automate execution.
Installation differs by platform.
Existing installs under paths like ~/.claude/abel-skills, ~/.codex/abel-skills, .claude/abel-skills, or .agents/abel-skills can keep their local folder names. To migrate an existing checkout, update its git remote:
git remote set-url origin https://github.com/Abel-ai-lab/PredictAnything.git
New install examples use PredictAnything.
Ask your AI agent to:
Note: if you are an AI agent reading this, suggest the above prompts for your human to get started.
A directional benchmark compared Abel Invest with an isolated LLM-only workflow on 1,000 tickers from 2020-01-01 to 2026-05-28. Abel Invest completed 997 / 1000 runs, while the no-skill workflow completed 962 / 1000; the comparable both-OK set covered 960 tickers.
| Measure | Abel Invest | No-skill LLM | Readout |
|---|---|---|---|
| OK coverage | 997 / 1000 (99.7%) | 962 / 1000 (96.2%) | Abel completed 35 more tickers |
| Mean Sharpe | 0.8245 | 0.2308 | 3.57x higher |
| Median Sharpe | 0.8139 | 0.2336 | 3.48x higher |
| Mean total return | 1.5221 | 0.6084 | 2.50x higher |
| Median total return | 1.0170 | 0.1393 | 7.30x higher |
| Median max drawdown | -0.1911 | -0.3306 | smaller typical drawdown |
| Mean return/drawdown | 7.4754 | 1.9765 | 3.78x higher |
| Median return/drawdown | 5.7227 | 0.5066 | 11.29x higher |
On the both-OK set, Abel Invest won on Sharpe (98.3%), total return (84.7%), max drawdown (79.3%, less negative is better), and return/drawdown (92.0%). Lower-tail behavior also improved: Abel Invest's 10th percentile Sharpe was positive at 0.5174, while the no-skill workflow was -0.2719.
This benchmark is directional capability evidence, not investment advice or a guarantee of live trading performance.
该项目提供了一个开源的AI工作流,适合量化代理和AI交易代理使用,提供了因果模型等技能,但需要进一步完善和测试。
该工具未明确声明开源协议,商业使用前请联系原作者确认授权范围,避免侵权风险。
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
经综合评估,量化AI和交易AI技能 在Agent工作流赛道中表现稳健,质量良好。如果你已有明确的使用需求,可以直接上手体验;如果还在评估阶段,建议对比同类工具后再做决策。
| 原始名称 | abel-strategy-research-skills |
| 原始描述 | 开源AI工作流:Quant AI and trading AI skills for quant agents and AI trading agents: causal ma。⭐9 · Python |
| Topics | workflowagent-skillsai-traderalgorithmic-tradingbacktestingbacktesting-trading-strategiespython |
| GitHub | https://github.com/Abel-ai-lab/abel-strategy-research-skills |
| 语言 | Python |
收录时间:2026-06-05 · 更新时间:2026-06-11 · License:未公布 · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端