Content
Tool List
**Agent Skills** repository for xtquant QMT / Research Terminal: Encapsulate domain knowledge for quantitative strategy development (factor backtesting, signal generation, live trading templates, etc.) into standard [SKILL.md](https://agentskills.io/specification) skills, which can be directly used in 70+ AI programming tools like Claude Code, Cursorx, Gemini CLI, Kimi Code, etc.
[](https://github.com/dfkai/xtquantai/actions/workflows/validate.yml)
[](LICENSE)
> This repository was previously the MCP server implementation of xtquant and has now been fully transformed into a skill repository. The old MCP code is preserved in the [`legacy-mcp`](https://github.com/dfkai/xtquantai/tree/legacy-mcp) branch (tag `v0.1.0-mcp`) and is no longer maintained.
## Skill List
| Skill | Status | Description |
|------|------|------|
| [`qmt-inner-backtest`](skills/qmt-inner-backtest/SKILL.md) | ✅ Available | Generate QMT built-in daily frequency cross-sectional factor backtest strategy based on strategy research report PDF interpretation of factor logic, using template script to generate signal and handlebar execution, including Barra style factor processing and anti-future function check |
| `qmt-future-trade` | 🚧 Planned | Futures open and close position strategy |
| `qmt-live-strategy-template` | 🚧 Planned | Target position live trading strategy template |
| `-live-signal-feishu` | 🚧 Planned | Signal generationishu notification |
## Community Participation
This project is driven by community needs - any repetitive tasks you encounter in QMT quantification may be worth turning into a skill:
- 💡 ** [Submit Skill Request](https://github.com/dfkai/xtquantai/issues/new?template=skill-request.yml), even a one-sentence idea is welcome
- 🐛 **Skill Issue** → [Report Bug](https://githubfkai/xtquantai/issues/new?template=bug-report.yml)
- 💬 **Usage Consultation / Quantization Idea Exchange** → [Discussions](https://github.com/dfkai/xtquantai/discussions)
🔧 **Contribute** → Read [Contribution Guide](CONTRIBUTING.md) and submitAll skills are automatically validated against the [Agent Skills standard](https://agentskills.io/specification) in CI. The maintenance process is documented in the [book](docs/maintenance-playbook.md).
## Installation
### Method 1: npx skills (Recommended, supports 70+ tools)
# Interactive selection of tools to install
npx skills add dfkai/xtquantai
# Install a specific skill
npx skills add dfkai/xtquantai --skill qmt-inner-backtest
# Install to global directory (available for all projects)
npx skills add dfkai/xtquantai -g
```
### Method 2: Claude Code Native Plugin
```text
/plugin marketplace add dfkai/xtquantai
/plugin install qmt-skills@xtquantai
```
### Method 3: Kimi Code CLI
```text
/plugins install https://github.com/dfkai/xtquantai
```
### Method 4: Let Your Agent Install
Tell any AI programming tool:
> Read https://raw.githubusercontent.com/dfkai/xtquantai/master/INSTALL.md and follow the instructions to install the skill.
### Method 5: Manual Copy
Copy the entire `skills/<skill-name>/` directory to your tool's skill directory:
| Tool | Project-level Directory | User-level Directory |
|------|-----------|-----------|
| Claude Code | `.claude/skills/` | `~/.claude/skills/` |
| Cursor (≥2.4) | `.cursor/skills/` (also reads `.agents/`, `.claude/`) | `~/.cursor/skills/` |
| OpenAI Codex CLI | `.agents/skills/` | `~/.agents/skills/` |
| GitHub Copilot | `.github/skills/` (also reads `.claude/`, `.agents/`) | — |
| Gemini CLI | `.gemini/` or `.agents/skills/` | — |
| Kimi Code CLI | `.kimi-code/skills/` or `.agents/skills/` | — |
| Windsurf | `.windsurf/skills/` or `.agents/skills/` | — |
| OpenCode | `.opencode/skills/`, `.claude/skills/`, `.agents/skills/` | — |
| Cline (≥3.48) | `.cline/skills/` or `.claude/skills/` (experimental features required) | — |
| Byte Trae | Only `.trae/skills/` | — |
| Qwen Code | Only `.qwen/skills/` | — |
| iFlow CLI | Only `.iflow/skills/` | — |
> The cross-tool generic directory is `.agents/sk; only Claude tools `.claude/skills/`.
## Usage Example
After installing `qmttest`, directly say in your AI tool:
> This (attached PDF/screenshot) has upper and lower shadow line factors, please help me generate a QMT backtest script for CSI 1000, Top 10 and 5-day rebalancing.
The Agent will first output a strategy specification table for your confirmation, then generate a complete QMT strategy editor backtest script based on, and prompt you to manually verify configuration items in the QMT panel.
⚠️ QMT / Research Terminal only supports Windows, and the generated strategy script needs to be run in the QMT strategy editor on Windows; the skill itself (script generation process) is platform-independent.
## Repository Structure
```
xtquantai/
├── skills/ # Skill directory (each subdirectory is a skill)
│ └── qmt-inner-backtest/
│ ├── SKILL.md # Skill definition (agentskills.io standard format)
│ └── scripts/
│ └── daily-factors-backtest.py # Backtest template script
├── .claude-plugin/
│ └── marketplace.json # Claude Code plugin market list
├── .github/ # Issue/PR templates and CI validation
├── kimi.plugin.json # Kimi Code plugin list
├── INSTALL.md # Installation instructions for agents
├── CONTRIBUTING.md # Contribution guide
└── docs/ # Research documents and maintenance manual
```
## License
MIT — See [LICENSE](LICENSE).
## Acknowledgements
- [Agent Skills](https://agentskills.io/) open skill standard
