Quantum-Practices is a DeepSeek Harness tool bundle for quantum algorithm best practices. It provides structured, reviewable guidance to DeepSeek Harness agents through a read-only model-facing tool.
As a DeepSeek Harness plugin, it registers one read-only quantum_practices tool for listing, searching, and reading packaged quantum algorithm practice guides from an immutable build-time catalog.
Quantum-Practices is based on and adapted from the GitHub project unitarylab/quantum-skills. The original project provides the quantum algorithm guide corpus; this repository reworks that foundation into a DeepSeek Harness tool bundle with a generated, read-only practice catalog.
- Progressive Disclosure — Root
SKILL.mdis lightweight; algorithm and simulator guides load only when needed. - DeepSeek Harness Tool Bundle —
quantum_practicesexposeslist,search, andgetwithout executing code. - Read-Only Runtime — No network, subprocess, filesystem writes, Python execution, credentials, or native code.
- Best-Practice Coverage — Primitives, linear systems, cryptography, Hamiltonian simulation, Schrodingerization, eigensolvers, gradients, quantum machine learning, state preparation, and quantum error correction.
- Multi-Simulator Support — UnitaryLab (recommended), Qiskit, and PennyLane, with clear selection rules.
- GitHub-Sourced Corpus — Practice guides are synchronized from the public GitHub upstream only.
- Education-Friendly — Suitable for concept explanation, circuit design, code review, and hands-on demos.
| Category | Algorithms |
|---|---|
| Primitives | Grover, QPE, Hadamard Test, Hadamard Transform, Amplitude Amplification, Amplitude Estimation |
| Linear Systems | HHL, LCU, AQC, VQLS, QSVT-QLSA, QFT, Quantum Signal Processing (QSP) |
| Cryptography | Shor's Algorithm, Discrete Logarithm, Simon's Algorithm |
| Hamiltonian Simulation | Cartan decomposition, Trotter, QDrift, Taylor Series, QSP |
| Schrodingerization | Advection, Heat (1D/2D) |
| Eigensolvers | NumPyEigensolver, VQD |
| Gradients | Parameter-shift, Finite-difference, Linear-combination, SPSA, Reverse-mode, QFI |
| Quantum Machine Learning | VQE, VQC, QAOA, QCBM, CVQNN, Fermi-Hubbard VQE |
| State Preparation | Mottonen, MPS, Multiplexer, Pauli, Superposition |
| Quantum Error Correction | qLDPC, CSS Codes, Hypergraph Product Codes |
| Simulator | When to Use | Platform |
|---|---|---|
| UnitaryLab (default) | Learning, algorithm demos, PDE workflows | Win / macOS / Linux |
| Qiskit | Noise models, IBM hardware workflows | Win / macOS / Linux |
| PennyLane | Differentiable hybrid optimization | Win / macOS / Linux |
quantum-practices/
|
+-- SKILL.md # Root practice index used by the catalog
+-- README.md
+-- package.json # DeepSeek Harness tool-bundle metadata
+-- cordis.patch.yml # Profile Bundle patch
+-- src/ # DSH plugin source
+-- lib/ # Built release artifact
|
+-- algorithms/ # Quantum algorithm skills
| +-- primitives/ # Grover, QPE, Hadamard test/transform, AA, AE
| +-- linear-systems/ # HHL, LCU, AQC, VQLS, QSVT-QLSA, QFT, QSP
| +-- cryptography/ # Shor, discrete logarithm, Simon
| +-- hamiltonian-simulation/ # Cartan, Trotter, QDrift, Taylor, QSP
| +-- schrodingerization/ # Advection and heat-equation workflows
| +-- eigensolvers/ # NumPyEigensolver, VQD
| +-- gradients/ # Parameter-shift, finite-diff, SPSA, reverse, QFI
| +-- quantum-machine-learning/ # VQE, VQC, QAOA, QCBM, CVQNN
| +-- state-preparation/ # Mottonen, MPS, multiplexer, Pauli, superposition
| +-- quantum-error-correction/ # qLDPC, CSS codes
|
+-- simulators/ # Simulator selection & installation guides
+-- unitarylab/ # Recommended simulator guide
+-- qiskit/
+-- pennylane/
For most users, install Quantum-Practices into the DeepSeek Harness profile you use, then ask your agent to consult Quantum-Practices before answering quantum algorithm questions.
If you use the Web UI:
npx @deepseek-ai/dsh@0.1.0-rc.6 plugin --profile web add \
github:unitarylab/quantum-practices#mainRestart DeepSeek Harness Web after installation.
If you use the headless CLI:
npx @deepseek-ai/dsh@0.1.0-rc.6 plugin --profile headless add \
github:unitarylab/quantum-practices#mainFor local development, install this checkout directly:
dsh plugin --profile web add "/path/to/quantum-practices"
dsh plugin --profile headless add "/path/to/quantum-practices"For review or release evidence, replace main with a pinned 40-character commit SHA.
Verify that the profile contains the inserted row:
dsh --profile headless --dump-config | \
rg "tool-quantum-practices|dsh-unitarylab-quantum-practices"Expected output:
# == dsh-unitarylab-quantum-practices
- id: tool-quantum-practices
name: dsh-unitarylab-quantum-practices
Run a functional test:
npx @deepseek-ai/dsh@0.1.0-rc.6 --profile headless \
"Use the quantum_practices tool to find the HHL practice guide and explain the required matrix constraints."After installation, users can ask naturally. The model should call quantum_practices in the background:
Use Quantum-Practices to review HHL before explaining the matrix constraints on A.
Before writing Grover code, check Quantum-Practices and list the common implementation pitfalls.
Use Quantum-Practices to compare quantum phase estimation and the quantum Fourier transform.
Consult Quantum-Practices and recommend a simulator for a variational quantum algorithm.
Check Quantum-Practices and explain how Trotter and QDrift differ for Hamiltonian simulation.
By default, get returns a brief, token-conscious view with the most relevant sections. The model should request detail="full" only when the user needs full implementation notes, complete examples, or debugging context.
Developers can also inspect the tool contract directly:
quantum_practices(action="list")
quantum_practices(action="search", query="HHL linear system")
quantum_practices(action="get", id="algorithms/linear-systems/hhl")
quantum_practices(action="get", query="Explain HHL matrix constraints")
quantum_practices(action="get", query="Implement HHL with a 2x2 example", detail="full")
The DSH plugin never executes algorithms/**/scripts/*.py and never installs or imports Python dependencies.
npm ci
npm run check
npm pack --dry-run --jsonnpm run build regenerates src/generated/skill-catalog.ts and compiles the committed lib/ release artifact.
Quantum-Practices does not ship a root requirements.txt, bundled wheels, or a Python runtime. Any Python setup belongs to the separate project where you choose to run generated examples; it is not part of the DeepSeek Harness plugin install path.
This repository source is licensed under the MIT License.
Quantum-Practices is a derivative adaptation of unitarylab/quantum-skills. See NOTICE for attribution details.
Quantum-Practices 是一个面向量子算法最佳实践的 DeepSeek Harness 工具包。它通过一个只读模型工具，为 DeepSeek Harness Agent 提供结构化、可审查的量子算法实践指南。
作为 DeepSeek Harness 插件，它注册一个只读 quantum_practices 工具，用于从构建期固化的 Practice Catalog 中列出、搜索和读取量子算法实践指南。
Quantum-Practices 基于 GitHub 项目 unitarylab/quantum-skills 进行二次创作。原项目提供了量子算法指南语料；本仓库在此基础上改造为 DeepSeek Harness 工具包，并生成只读的 Practice Catalog。
- 渐进式加载 — 根
SKILL.md轻量，算法与模拟器指南仅在需要时才加载。 - DeepSeek Harness 工具包 —
quantum_practices提供list、search、get，不执行代码。 - 只读运行时 — 无网络、无 subprocess、无写盘、无 Python 执行、无 credentials、无 native code。
- 最佳实践覆盖 — 基元、线性系统、密码学、哈密顿量模拟、Schrodingerization、本征求解器、梯度方法、量子机器学习、态制备与量子纠错一应俱全。
- 多模拟器支持 — UnitaryLab（推荐）、Qiskit、PennyLane，附明确选型规则。
- GitHub 来源语料 — Practice guides 仅从公开 GitHub 上游同步。
- 教学友好 — 适用于概念解释、电路设计、代码审查和动手实验。
| 分类 | 算法 |
|---|---|
| 基础量子算法 | Grover、QPE、Hadamard 测试、Hadamard 变换、振幅放大、振幅估计 |
| 线性系统 | HHL、LCU、AQC、VQLS、QSVT-QLSA、QFT、量子信号处理（QSP） |
| 密码学 | Shor 算法、离散对数、Simon 算法 |
| 哈密顿量模拟 | Cartan 分解、Trotter、QDrift、Taylor 级数、QSP |
| Schrodingerization | 对流、热方程（一维/二维） |
| 本征求解器 | NumPyEigensolver、VQD |
| 梯度方法 | 参数位移、有限差分、线性组合、SPSA、反向模式、QFI |
| 量子机器学习 | VQE、VQC、QAOA、QCBM、CVQNN、Fermi-Hubbard VQE |
| 态制备 | Mottonen、MPS、Multiplexer、Pauli、Superposition |
| 量子纠错 | qLDPC、CSS 码、超图乘积码 |
| 模拟器 | 适用场景 | 平台 |
|---|---|---|
| UnitaryLab (默认) | 学习、算法演示、PDE 工作流 | Win / macOS / Linux |
| Qiskit | 噪声模型、IBM 硬件工作流 | Win / macOS / Linux |
| PennyLane | 可微分混合优化 | Win / macOS / Linux |
quantum-practices/
|
+-- SKILL.md # Catalog 使用的根实践索引
+-- README.md
+-- package.json # DeepSeek Harness tool-bundle 元数据
+-- cordis.patch.yml # Profile Bundle patch
+-- src/ # DSH 插件源码
+-- lib/ # 编译后的 release artifact
|
+-- algorithms/ # 量子算法技能
| +-- primitives/ # Grover、QPE、Hadamard 测试/变换、振幅放大与估计
| +-- linear-systems/ # HHL、LCU、AQC、VQLS、QSVT-QLSA、QFT、QSP
| +-- cryptography/ # Shor、离散对数、Simon
| +-- hamiltonian-simulation/ # Cartan、Trotter、QDrift、Taylor、QSP
| +-- schrodingerization/ # 对流与热方程工作流
| +-- eigensolvers/ # NumPyEigensolver、VQD
| +-- gradients/ # 参数位移、有限差分、SPSA、反向模式、QFI
| +-- quantum-machine-learning/ # VQE、VQC、QAOA、QCBM、CVQNN
| +-- state-preparation/ # Mottonen、MPS、Multiplexer、Pauli、Superposition
| +-- quantum-error-correction/ # qLDPC、CSS 码
|
+-- simulators/ # 模拟器选型与安装指南
+-- unitarylab/ # 推荐模拟器指南
+-- qiskit/
+-- pennylane/
普通用户不需要理解底层 action。把 Quantum-Practices 安装进正在使用的 DeepSeek Harness profile 之后，直接让 Agent 先查 Quantum-Practices，再回答量子算法问题即可。
如果你使用 Web UI：
npx @deepseek-ai/dsh@0.1.0-rc.6 plugin --profile web add \
github:unitarylab/quantum-practices#main安装后重启 DeepSeek Harness Web。
如果你使用 headless CLI：
npx @deepseek-ai/dsh@0.1.0-rc.6 plugin --profile headless add \
github:unitarylab/quantum-practices#main本地开发时，可以直接安装当前 checkout：
dsh plugin --profile web add "/path/to/quantum-practices"
dsh plugin --profile headless add "/path/to/quantum-practices"审核或 release evidence 建议把 main 换成固定的 40 位 commit SHA。
验证 profile 中是否出现插入的 row：
dsh --profile headless --dump-config | \
rg "tool-quantum-practices|dsh-unitarylab-quantum-practices"期望输出：
# == dsh-unitarylab-quantum-practices
- id: tool-quantum-practices
name: dsh-unitarylab-quantum-practices
做一次真实功能测试：
npx @deepseek-ai/dsh@0.1.0-rc.6 --profile headless \
"Use the quantum_practices tool to find the HHL practice guide and explain the required matrix constraints."安装完成后，用户可以直接自然提问；模型应在后台调用 quantum_practices：
请先查 Quantum-Practices，再解释 HHL 对矩阵 A 的约束。
写 Grover 代码前，请查 Quantum-Practices 并列出常见实现错误。
请根据 Quantum-Practices 比较量子相位估计和量子傅里叶变换。
请查 Quantum-Practices，并建议变分量子算法应该使用哪个 simulator。
请参考 Quantum-Practices，说明 Trotter 和 QDrift 在哈密顿量模拟中的区别。
默认情况下，get 返回省 token 的 brief 视图，只包含最相关的章节。只有当用户需要完整实现说明、完整示例或调试上下文时，模型才应该请求 detail="full"。
开发者也可以直接查看工具接口：
quantum_practices(action="list")
quantum_practices(action="search", query="HHL linear system")
quantum_practices(action="get", id="algorithms/linear-systems/hhl")
quantum_practices(action="get", query="Explain HHL matrix constraints")
quantum_practices(action="get", query="Implement HHL with a 2x2 example", detail="full")
DSH 插件不会执行 algorithms/**/scripts/*.py，也不会安装或导入 Python 依赖。
npm ci
npm run check
npm pack --dry-run --jsonnpm run build 会重新生成 src/generated/skill-catalog.ts，并编译需要提交的 lib/ release artifact。
Quantum-Practices 不发布根 requirements.txt、内置 wheel 或 Python runtime。任何 Python 环境都应属于你实际运行示例的独立项目，不属于 DeepSeek Harness 插件安装路径。
本仓库源码采用 MIT License。
Quantum-Practices 是基于 unitarylab/quantum-skills 的二次创作。来源说明详见 NOTICE。