math-skill 2.0.1 → 3.0.0

package/lenses/spectral.en.md

@@ -0,0 +1,52 @@
+# 📡 Spectral Decomposition Lens
+
+> Any linear operator can be decomposed into a superposition of eigencomponents — eigenvalues reveal dominant structure, eigenvectors reveal dominant directions.
+
+## What Perspective It Offers
+
+This is a "decomposer's" perspective — breaking complex linear operators (matrices, kernel functions, graph Laplacians) into spectra of eigenvalues and singular values, using dominant eigencomponents to capture global behavior and tail eigenvalues to quantify noise and redundancy. The core conviction: the spectrum — the distribution of eigenvalues — determines everything about an operator: its condition number, rank, stability, and convergence rate.
+
+## What Problems It Is Suited to Diagnose
+
+- What is the dominant structure of a matrix or operator? What is its effective rank?
+- Optimizer convergence is slow — is the condition number too large? How should a preconditioner be designed?
+- Where are the critical bottlenecks in a graph or network? Is the spectral gap too small?
+- Need to reduce dimensions while preserving maximum variance — are the first k principal components sufficient?
+- How large is the spectral drift after quantization or pruning? Has stability been compromised?
+
+## What Problems It Is Not Suited For
+
+- Problems dominated by nonlinear effects — spectral decomposition is a linear tool
+- Operators that are not diagonalizable (non-normal matrices, defective matrices) — Jordan form is required
+- Problems that require preserving global topological structure — spectral methods may destroy topology
+
+## Which Knowledge Domains It Routes To
+
+- **matrix-analysis/spectral-decomposition**: EVD, Schur decomposition, normal matrices
+- **matrix-analysis/low-rank-approximation**: Eckart-Young theorem, randomized SVD
+- **matrix-analysis/matrix-perturbation**: Weyl's inequality, Davis-Kahan theorem
+- **matrix-analysis/positive-semidefinite**: Spectral properties of PSD matrices, Cholesky factorization
+
+## What AI Designs It May Inspire
+
+- **Spectral Normalization**: Constrain the largest singular value of weight matrices to stabilize training
+- **Low-Rank Adaptation (LoRA)**: Fine-tune in a truncated SVD subspace
+- **Spectral Graph Attention**: Use graph Laplacian eigenvectors for positional encoding
+- **Spectral Clustering Routing**: Use top-k eigenvectors for MoE token assignment
+- **Effective Rank Monitoring**: Track the stable rank of weight matrices to detect overfitting
+
+## Reasoning Protocol
+
+1. **Identify the operator**: Which matrix, kernel, or graph requires spectral analysis? What are its dimensions?
+2. **Compute or estimate the spectrum**: Full EVD (small matrices) / power iteration for dominant eigenvalues / randomized SVD
+3. **Analyze the spectral distribution**: Condition number, spectral gap, effective rank, decay rate
+4. **Identify dominant components**: What proportion of the variance or Frobenius norm is explained by the first k eigenvalues?
+5. **Assess truncation impact**: Quantify the error introduced by discarding tail eigenvalues (Eckart-Young provides a sharp bound)
+
+## Acceptance Criteria
+
+- The target operator is clearly defined
+- The spectrum (or dominant eigenvalues/singular values) has been computed or estimated
+- Spectral distribution metrics are reported (condition number, spectral gap, effective rank)
+- Truncation and approximation decisions are backed by quantitative error bounds (Weyl / Davis-Kahan / Eckart-Young)
+- The output includes actionable conclusions (preconditioner design, dimensionality selection, stability guarantees)

package/lenses/spectral.md

@@ -0,0 +1,52 @@
+# 📡 谱分解透镜 / Spectral Decomposition Lens
+
+> 任何线性算子都可以分解为特征分量的叠加——特征值揭示主导结构，特征向量揭示主导方向。
+
+## 它是什么视角
+
+这是一种"分解者"的视角——将复杂线性算子（矩阵、核函数、图 Laplacian）分解为特征值/奇异值的谱，用主导特征分量捕获全局行为，用尾部特征量化噪声与冗余。核心信念：谱（特征值分布）决定了算子的一切——条件数、秩、稳定性、收敛速度。
+
+## 适合诊断什么问题
+
+- 矩阵/算子的主导结构是什么？有效秩是多少？
+- 优化器收敛速度慢——条件数是否过大？预条件如何设计？
+- 图/网络的关键瓶颈在哪里？谱隙（spectral gap）是否过小？
+- 需要降维但保留最大方差——前 k 个主成分够不够？
+- 量化/剪枝后的谱漂移有多大？稳定性是否被破坏？
+
+## 不适合什么问题
+
+- 非线性效应主导的问题——谱分解是线性工具
+- 算子本身不可对角化（非正规矩阵、亏损矩阵）——需要 Jordan 形式
+- 需要保留全局拓扑结构的问题——谱方法可能破坏拓扑
+
+## 会路由到哪些知识域
+
+- **matrix-analysis/spectral-decomposition**：EVD、Schur 分解、正规矩阵
+- **matrix-analysis/low-rank-approximation**：Eckart-Young 定理、随机化 SVD
+- **matrix-analysis/matrix-perturbation**：Weyl 不等式、Davis-Kahan 定理
+- **matrix-analysis/positive-semidefinite**：PSD 矩阵的谱性质、Cholesky
+
+## 可能产生哪些 AI 设计
+
+- **谱归一化**：约束权重矩阵的最大奇异值，稳定训练
+- **低秩适配器 (LoRA)**：在截断 SVD 子空间中做微调
+- **谱图注意力**：用图 Laplacian 特征向量做位置编码
+- **谱聚类路由**：用 top-k 特征向量做 MoE token 分配
+- **有效秩监控**：跟踪权重矩阵的稳定秩，检测过拟合
+
+## 推理协议
+
+1. **识别算子**：哪个矩阵/核/图需要谱分析？维度多大？
+2. **计算或估计谱**：完整 EVD（小矩阵）/ 幂迭代估计主导特征值 / 随机化 SVD
+3. **分析谱分布**：条件数、谱隙、有效秩、衰减率
+4. **识别主导分量**：前 k 个特征值解释了多大比例的方差/Frobenius 范数？
+5. **评估截断影响**：丢弃尾部特征造成的误差（Eckart-Young 给出精确界）
+
+## 验收标准
+
+- 目标算子已明确定义
+- 谱（或主导特征值/奇异值）已计算或估计
+- 谱分布指标已给出（条件数、谱隙、有效秩）
+- 截断/近似决策有量化误差界（Weyl/Davis-Kahan/Eckart-Young）
+- 输出包含可执行结论（预条件设计、降维维度选择、稳定性保证）

package/lenses/symmetry.en.md

@@ -0,0 +1,49 @@
+# ⚛️ Symmetry Lens
+
+> Seek the invariant amid change — every symmetry corresponds to a conserved quantity, and every invariant is a key to simplifying the problem
+
+## What Perspective It Offers
+
+Symmetry is a way of "finding what remains unchanged under transformations": when confronting a complex system, first ask, "Under what transformations do which properties of the system remain invariant?" Invariants are the core tool for simplification — they distinguish only between orbits (equivalence classes), not between elements within the same orbit. A complete set of invariants can embed the entire space into a simpler quotient space, achieving a full classification. Noether's theorem tells us that every continuous symmetry necessarily corresponds to a conservation law.
+
+## What Problems It Is Suited to Diagnose
+
+- Facing a complex system and seeking simplification clues — use invariants to reduce dimensionality
+- Needing to classify or identify objects — use orbits (the quotient space X/G)
+- Searching for conserved quantities or invariants — functions that are constant on orbits
+- Determining the solvability of equations (Galois theory) or designing equivariant structures
+
+## What Problems It Is Not Suited For
+
+- The system is completely asymmetric with no discernible pattern — no group action is available to exploit
+- Precise numerical solutions are required — symmetry provides structural information, not specific values
+- Symmetry breaking is the central mechanism — one should turn to analyzing the breaking pattern rather than searching for invariants
+- The group structure is too complex — when the quotient space is no simpler than the original space
+
+## Which Knowledge Domains It Routes To
+
+- `lie-theory/group-action`: Group actions, the orbit-stabilizer theorem, and Burnside's lemma — the algebraic foundations of symmetry
+- **lie-theory**: Lie groups, Lie algebras, and Noether's theorem — continuous symmetries and conservation laws
+- `lie-theory/lie-group`: Galois groups and solvable groups — algebraic criteria for the solvability of equations
+
+## What AI Designs It May Inspire
+
+- **Symmetry Detector**: Identifies candidate transformation groups from a problem description and verifies the group axioms
+- **Invariant Extractor**: Computes invariants using the Reynolds operator or Lie algebra generators
+- **Quotient-Space Classifier**: Works on X/G and uses a complete set of invariants to classify equivalence classes
+
+## Reasoning Protocol
+
+1. **Identify the Transformation Group**: List the candidate transformations of the system, organize them into a candidate group, and verify closure, associativity, identity, and inverses
+2. **Find Invariants**: For finite groups, use the Reynolds operator; for continuous groups, use Lie algebra generator equations
+3. **Simplify Using Invariants**: Work on the quotient space X/G, replacing constrained variables with invariants
+4. **Orbit Classification**: Use the orbit-stabilizer theorem and assess the completeness of the invariant set
+5. **Check for Symmetry Breaking**: Identify G → H breaking patterns and analyze Goldstone modes
+
+## Acceptance Criteria
+
+- The candidate group has been verified through the axioms (not merely assumed to have group structure)
+- Invariants have been explicitly given, distinguishing orbital invariants from global invariants
+- The simplification strategy has stated the degree of dimensionality reduction
+- Symmetry breaking has been checked (presence or absence; spontaneous or explicit)
+- Classification results and the completeness of the invariant set have been assessed

package/lenses/symmetry.md

@@ -0,0 +1,49 @@
+# ⚛️ 对称透镜
+
+> 在变化中寻找不变——每一个对称性都对应一个守恒量，每一个不变量都是简化问题的钥匙
+
+## 它是什么视角
+
+对称是一种"在变换下寻找不变"的眼光：面对复杂系统时，先问"什么变换下系统的哪些性质不变？"不变量是简化问题的核心工具——它只区分轨道（等价类），不区分同轨道内的元素。完备的不变量集能将整个空间嵌入更简单的商空间，实现完全分类。Noether 定理告诉我们：连续对称必然对应守恒律。
+
+## 适合诊断什么问题
+
+- 面对复杂系统想找简化线索——利用不变量降维
+- 需要分类/识别对象——利用轨道（商空间 X/G）
+- 寻找守恒量或不变量——轨道上取常值的函数
+- 判断方程可解性（Galois 理论）或设计等变结构
+
+## 不适合什么问题
+
+- 系统完全不对称且无规律——无群作用可利用
+- 需要精确数值解——对称性给结构信息而非具体数值
+- 对称破缺是核心机制——应转向分析破缺模式而非找不变量
+- 群结构过于复杂——商空间不比原空间简单时
+
+## 会路由到哪些知识域
+
+- `lie-theory/group-action`：群作用、轨道-稳定子定理、Burnside 引理——对称性的代数基础
+- **lie-theory**：李群/李代数、Noether 定理——连续对称与守恒律
+- `lie-theory/lie-group`：Galois 群、可解群——方程可解性的代数判据
+
+## 可能产生哪些 AI 设计
+
+- **对称性探测器**：从问题描述中识别候选变换群，验证群公理
+- **不变量提取器**：利用 Reynolds 算子或 Lie 代数生成元计算不变量
+- **商空间分类器**：在 X/G 上工作，用完备不变量集实现等价类分类
+
+## 推理协议
+
+1. **识别变换群**：列出系统可能的变换，组织为候选群，验证封闭/结合/单位/逆
+2. **寻找不变量**：有限群用 Reynolds 算子，连续群用 Lie 代数生成元方程
+3. **利用不变量简化**：在商空间 X/G 上工作，用不变量替代被约束变量
+4. **轨道分类**：利用轨道-稳定子定理，评估不变量完备性
+5. **检查对称破缺**：识别 G→H 破缺模式，分析 Goldstone 模数
+
+## 验收标准
+
+- 候选群已通过公理验证（不是假设它有群结构）
+- 不变量已显式给出，区分了轨道不变量与全局不变量
+- 简化策略已说明维度缩减程度
+- 对称破缺已检查（存在/破缺、自发/显式）
+- 分类结果与不变量完备性已评估

package/lenses/topological.en.md

@@ -0,0 +1,48 @@
+# 🌀 Topological Lens
+
+> Stretching and bending are allowed, but tearing is not — the truly important properties are those invariant under continuous deformation
+
+## What Perspective It Offers
+
+Topology is a way of "ignoring precise measurements and focusing only on qualitative structure": when exact distances do not matter, topology captures the essential structure — connectedness, the number of holes, and dimension. A donut and a coffee cup are topologically equivalent because each has exactly one hole. The core insight is that shape matters more than size; invariants that survive continuous deformation are the truly robust properties. Cohomology (especially Čech cohomology) can also detect whether "local fragments can be assembled into a globally consistent object."
+
+## What Problems It Is Suited to Diagnose
+
+- Qualitative classification — not "how large" or "how far," but "is it connected?" and "how many holes does it have?"
+- Robustness analysis — topological invariants guarantee that properties do not vanish under continuous perturbation
+- Data whose shape standard statistics cannot capture — clusters, voids, and ring-like structures
+- Detecting global consistency obstructions — whether local fragments can be assembled into a globally consistent object
+
+## What Problems It Is Not Suited For
+
+- Precise measurements are required — topology concerns only qualitative structure and does not provide distances or angles
+- Metric properties are central — topologically equivalent spaces can have entirely different metric behavior
+- Precise numerical decisions are needed — the topological perspective is too coarse to answer "how much" questions
+
+## Which Knowledge Domains It Routes To
+
+- `topology/persistent-homology`: Fundamental group, homology groups, Betti numbers, and cohomology — computational tools for topological invariants
+- **tda**: Persistent homology, filtrations, and simplicial complexes — extracting topological features from data
+- `topology/fundamental-group`: Sheaf cohomology and Čech cohomology — algebraic criteria for local-to-global consistency
+
+## What AI Designs It May Inspire
+
+- **Topological Feature Extractor**: Builds filtrations from point-cloud data and computes persistent homology barcodes
+- **Consistency Detector**: Uses Čech cohomology H¹ to detect obstructions to global consistency among local fragments
+- **Shape Classifier**: Uses combinations of invariants (χ, π₁, βₙ) to perform topological classification of spaces
+
+## Reasoning Protocol
+
+1. **Specify the Equivalence Standard**: Determine whether the relevant notion is homeomorphism, homotopy equivalence, or diffeomorphism; distinguish what may vary (distance) from what must remain invariant (number of holes)
+2. **Compute Topological Invariants**: Euler characteristic, connected components, fundamental group, homology groups, and Betti numbers
+3. **Classify Using Invariants**: Same χ implies same surface type; same π₁ implies same homotopy type
+4. **Construct a Topological Model**: Build a filtration from a point cloud, a graph from a network, or a phase space from a dynamical system
+5. **Verify Equivalence**: Attempt to construct an explicit homeomorphism or homotopy; check whether all invariants match
+
+## Acceptance Criteria
+
+- The equivalence standard has been explicitly stated (homeomorphism, homotopy equivalence, or diffeomorphism)
+- Invariants have been computed: at minimum, χ, π₁, and βₙ (where obtainable) have been provided
+- A classification conclusion has been given, with a note on whether the invariants are complete
+- Equivalence verification has been completed (match, mismatch, or inconclusive)
+- The reasoning chain is complete: "because [topological property] → [conclusion]"

package/lenses/topological.md

@@ -0,0 +1,48 @@
+# 🌀 拓扑透镜
+
+> 可以拉伸弯曲但不能撕裂——真正重要的性质在连续变形中不变
+
+## 它是什么视角
+
+拓扑是一种"忽略精确度量、只看定性结构"的眼光：当精确距离不重要时，拓扑捕获本质结构——连通性、洞的个数、维数。甜甜圈与咖啡杯拓扑等价，因为都恰好有一个洞。核心洞察：形状比大小重要，连续变形下存活的不变量才是真正鲁棒的性质。上同调（特别是 Cech 上同调）还能检测"局部片段能否拼成全局相容对象"。
+
+## 适合诊断什么问题
+
+- 需要定性分类——不在乎"多大""多远"，只在乎"是否连通""有多少洞"
+- 需要鲁棒性分析——拓扑不变量保证性质在连续扰动下不消失
+- 数据有标准统计无法捕获的形状——聚类、空洞、环状结构
+- 检测全局一致性障碍——局部片段能否拼成全局相容对象
+
+## 不适合什么问题
+
+- 需要精确测量——拓扑只关心定性结构，不给距离/角度
+- 度量性质是核心——拓扑等价空间可有完全不同的度量行为
+- 需要精确数值决策——拓扑视角过粗，回答不了"多少"的问题
+
+## 会路由到哪些知识域
+
+- `topology/persistent-homology`：基本群、同调群、Betti 数、上同调——拓扑不变量的计算工具
+- **tda**：持续同调、滤流、单纯复形——从数据中提取拓扑特征
+- `topology/fundamental-group`：层上同调、Cech 上同调——从局部到全局的一致性代数判据
+
+## 可能产生哪些 AI 设计
+
+- **拓扑特征提取器**：从数据点云构建滤流，计算持续同调条形码
+- **一致性检测器**：用 Cech 上同调 H^1 检测局部片段的全局相容性障碍
+- **形状分类器**：利用不变量组合（chi, pi_1, beta_n）对空间做拓扑分类
+
+## 推理协议
+
+1. **明确等价标准**：确定同胚/同伦/微分同胚，区分什么可变（距离）什么不变（洞数）
+2. **计算拓扑不变量**：欧拉示性数、连通分量、基本群、同调群、Betti 数
+3. **用不变量分类**：同 chi 同类型曲面；同 pi_1 同同伦类型
+4. **构造拓扑模型**：点云建滤流/网络建图/系统建相空间
+5. **验证等价性**：尝试构造显式同胚/同伦，检查不变量是否全部匹配
+
+## 验收标准
+
+- 等价标准已明确（同胚/同伦/微分同胚）
+- 不变量已计算：chi, pi_1, beta_n 至少给出可得的
+- 分类结论已给出，标注了不变量是否完备
+- 等价验证已完成（匹配/不匹配/不确定）
+- 推理链条完整："因为[拓扑性质] → [结论]"

package/lenses/variational.en.md

@@ -0,0 +1,48 @@
+# ⚖️ Variational Lens
+
+> Pursue the best among the feasible under constraints — convexity determines difficulty, duality reveals structure
+
+## What Perspective It Offers
+
+The variational perspective (formerly "Optimization") is a way of viewing "rational decision-making under constraints": any decision problem can be formulated as maximizing or minimizing an objective subject to constraints. The core idea is not to pursue an abstract "best," but to find the optimum within the feasible region. Convexity is the critical watershed — for convex problems, a local optimum is the global optimum; for non-convex problems, one must guard against local traps. Shadow prices (Lagrange multipliers) reveal which constraints truly matter.
+
+## What Problems It Is Suited to Diagnose
+
+- Resource allocation and trade-off decisions — maximizing or minimizing a quantifiable objective under constraints
+- Convexity determination — assessing problem difficulty and selecting solution strategies
+- Duality structure analysis — when the primal problem is hard, the dual may be more tractable
+- Sensitivity analysis — how the optimal solution responds to changes in constraints
+
+## What Problems It Is Not Suited For
+
+- Problems with no clear evaluation criterion — define the objective before optimizing
+- Purely executional tasks — there is no room for optimization
+- Problems that are inherently qualitative rather than quantitative extrema — model first, then optimize
+
+## Which Knowledge Domains It Routes To
+
+- `optimization/convex-optimization`: Convexity determination, properties of convex sets and functions — determines the difficulty class of the problem
+- **optimization**: Solution methods for LP, QP, convex, non-convex, combinatorial, and stochastic optimization
+- **matrix-analysis**: Condition numbers, preconditioning, and low-rank approximation — feasibility analysis for second-order methods
+
+## What AI Designs It May Inspire
+
+- **Objective–Constraint Decomposer**: Automatically extracts the objective function, hard and soft constraints, and the feasible region from a problem description
+- **Dual-Solution Router**: Detects whether the primal or dual problem is more tractable and selects the optimal solution path
+- **Sensitivity Monitor**: Tracks active constraints and shadow prices, triggering re-optimization when constraints change
+
+## Reasoning Protocol
+
+1. **Define the Objective**: Specify what to maximize or minimize; determine convexity (if convex, local = global)
+2. **List Constraints**: Distinguish hard from soft constraints, inequality from equality, linear from nonlinear
+3. **Classify Problem Type**: LP, QP, convex, non-convex, combinatorial, or stochastic — match to an appropriate solution method
+4. **Solve and Verify**: Exploit convexity or employ global search; check feasibility and optimality
+5. **Sensitivity Analysis**: Compute shadow prices, identify active constraints, and assess robustness
+
+## Acceptance Criteria
+
+- The objective function and its convexity status have been explicitly annotated
+- All constraints (including implicit ones) have been exhaustively listed
+- The problem type has been classified and a solution method has been matched
+- Sensitivity analysis has been provided (at minimum, active constraints are identified)
+- Multi-objective scenarios have been subjected to Pareto analysis

package/lenses/variational.md

@@ -0,0 +1,48 @@
+# ⚖️ 变分透镜
+
+> 在约束下追求可行中的最优——凸性决定难度，对偶揭示结构
+
+## 它是什么视角
+
+变分（原"优化"）是一种"约束下的理性决策"眼光：任何决策问题都可以表述为在约束条件下最大化或最小化某个目标。核心不是追求抽象的"最好"，而是在可行域中寻找最优。凸性是关键分水岭——凸问题局部最优即全局最优，非凸问题则需警惕局部陷阱。影子价格（拉格朗日乘子）揭示哪些约束真正重要。
+
+## 适合诊断什么问题
+
+- 资源分配与取舍决策——在约束下最大化/最小化可量化目标
+- 凸性判定——判断问题难度，选择求解策略
+- 对偶结构分析——原问题难解时，对偶问题可能更易处理
+- 灵敏度分析——约束变化时最优解如何响应
+
+## 不适合什么问题
+
+- 没有明确评价标准——先定义目标再优化
+- 纯执行性任务——没有优化空间
+- 本质是定性判断而非定量极值——应先建模再优化
+
+## 会路由到哪些知识域
+
+- `optimization/convex-optimization`：凸性判定、凸集/凸函数性质——决定问题难度等级
+- **optimization**：LP/QP/凸/非凸/组合/随机优化的求解方法
+- **matrix-analysis**：条件数、预条件、低秩近似——二阶法的可行性分析
+
+## 可能产生哪些 AI 设计
+
+- **目标-约束分解器**：从问题描述中自动提取目标函数、硬/软约束、可行域
+- **对偶求解路由**：检测原问题与对偶问题哪个更易解，选择最优路径
+- **灵敏度监控器**：跟踪活跃约束与影子价格，约束变化时触发重优化
+
+## 推理协议
+
+1. **定义目标**：明确最大化/最小化什么，判断凸性（凸则局部=全局）
+2. **列出约束**：区分硬/软约束、不等式/等式、线性/非线性
+3. **分类问题类型**：LP/QP/凸/非凸/组合/随机，匹配求解方法
+4. **求解与验证**：利用凸性或全局搜索求解，检查可行性与最优性
+5. **灵敏度分析**：计算影子价格，识别活跃约束，评估鲁棒性
+
+## 验收标准
+
+- 目标函数与凸性已明确标注
+- 所有约束（含隐式约束）已穷尽列出
+- 问题类型已分类，求解方法已匹配
+- 灵敏度分析已给出（至少标注活跃约束）
+- 多目标场景已做帕累托分析

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "math-skill",
-  "version": "2.0.1",
-  "description": "数学思想武器 v2：自动触发的现代数学能力激活器，把代数几何/微分几何/李理论等现代数学激活到算法与 GPU 协同设计研究中——既在 math 上 beautiful，又 GPU friendly。Mathematical Thinking Weapons v2: an auto-triggered activator that brings modern mathematics (algebraic geometry, differential geometry, Lie theory, …) into algorithm & GPU co-design research — beautiful in math, friendly to GPUs.",
+  "version": "3.0.0",
+  "description": "面向 AI 架构创新的数学研究操作系统 v3：三层正交架构（思想透镜 / 数学知识库 / 设计翻译层），自动诊断用户意图并路由。Math Research OS for AI architecture design: thinking lenses + math knowledge base + design translation layer with auto intent diagnosis.",
   "keywords": [
     "math",
     "mathematics",
@@ -41,7 +41,12 @@
     "agentic-workflow",
     "auto-trigger",
     "co-design",
-    "research"
+    "research",
+    "bilingual",
+    "lenses",
+    "knowledge-base",
+    "design-patterns",
+    "math-research-os"
   ],
   "license": "MIT",
   "author": "the-thinker0",
@@ -55,9 +60,12 @@
   },
   "files": [
     "README.md",
+    "README.en-US.md",
     "LICENSE",
     "commands/",
     "skills/",
+    "lenses/",
+    "design-patterns/",
     "agents/",
     "knowledge-base/",
     "references/"

package/references/agentic-workflow.en.md

@@ -0,0 +1,53 @@
+# Agentic Research Workflow / Human-in-the-Agent-Loop
+
+> Distilled from frontline auto-research practice (Infra / Algorithms / Chip Co-design). It defines the **collaborative mindset** this skill expects: what humans do, what the agent does, and what loop grounds "activating mathematical capability" in practice.
+> The activator references this file at runtime. This file defines the collaboration style the skill expects.
+
+## Creed
+
+1. **The mathematical foundation of this AI revolution is the first time 20th-century mathematics has stepped onto the commercial computing stage** — category theory, algebraic topology, algebraic geometry. The mathematical foundations of most mainstream algorithms still largely rest on the calculus and linear algebra of 1800–1900. Activating modern mathematics in algorithm design is the most important task during the algorithm exploration phase.
+2. **Humility**: The breadth of knowledge from the pre-training stage already surpasses any individual. **More knowledge is learned from the model than taught to it.** Letting go of the bias toward "teaching the model" usually yields better results.
+3. **Mathematical beauty ≠ computability**: Many mathematically elegant constructs break down in the face of GPU parallelism and low-precision arithmetic errors. Every deliverable must pass the acceptance gate defined in `gpu-friendly-math.md`.
+
+## Division of Labor
+
+**Humans are responsible for (what humans excel at and find most valuable):**
+- **Cross-domain activation**: Transferring the structure of one domain to another. The model already contains sufficient internal knowledge, but "without someone to activate it, that knowledge is unlikely to be developed." Example: applying algebraic geometry perspectives to study model architectures; historically, the correspondence between fiber bundles and gauge fields — the structure was already there, awaiting only a single cross-domain activation.
+- **Avoid hard-coding subjective bias**: The more rules you add, the more you constrain the model's ability to explore alternatives. To get good results, **do not impose human subjective preferences by force.**
+- **Expanding cognitive context with the model**: The human brain can manually compare at most 2–3 variables simultaneously, while the model's long context window can weigh many at once. Leveraging the model to expand an individual's cognitive context is a core principle of agentic development.
+- **Choosing directions and setting acceptance gates**: Deciding which path to explore and what standards to apply for acceptance.
+
+**The Agent is responsible for (what the model excels at):**
+- **Large-context search and enumeration**: Enumerating large numbers of candidate structures, comparing multi-parameter combinations in parallel.
+- **Verification**: Checking mathematical correctness + GPU feasibility.
+- **Building and executing test plans**: Recording research tasks as trackable markdown tables, autonomously adding new cases, fixing compilation errors, and retrying on failure.
+
+## The Activation Loop
+
+> The activator's `Diagnosis → Mapping → GPU Filtering` main pipeline is a distillation of this loop.
+
+1. **Diagnosis**: What is the algorithmic structure / bottleneck of the problem? (Do not start by piling on math tutorials.)
+2. **Cross-domain activation**: Scan `books/*.md` for transferable modern mathematical structures, **enumerate multiple candidates** (leveraging the model's large-context advantage).
+3. **Dual acceptance gate**: Each candidate passes two gates — (1) Mathematically correct (differentiable, self-consistent, with correctness guarantees); (2) GPU-feasible (the eight dimensions of `gpu-friendly-math.md`).
+4. **Iterative tracking**: Record candidates, scores, and status in a markdown test plan, converging step by step.
+
+## Markdown Test Plan Template / Tracking Template
+
+Record research tasks in a table that the Agent can continuously update:
+
+```markdown
+| Candidate Structure | Math Correctness | GPU 8-Dim | Complexity | Status | Notes / Next Steps |
+|--------------------|-----------------|-----------|------------|--------|-------------------|
+| Tropical gating replacing TopK | ✅ Differentiable relaxation | 1✅2❌3✅… | Sub-quadratic (per-token) | Verifying | Compare with SWA baseline |
+| …                  |                 |           |            | Todo   |                   |
+```
+
+The Agent is responsible for tracking execution and updating this table; the human is responsible for reading the table and adjusting direction.
+
+## Operating Principles
+
+- **Diagnose first, then map, then filter** — do not produce lengthy math tutorials; go straight to structural mapping and GPU feasibility.
+- **Preserve exploratory freedom**: Add fewer hard rules; leave room for the model to explore.
+- **Humans should be "lazy"**: Delegate all tedious work to the Agent; humans focus only on cross-domain activation and directional judgment.
+- **Progressive disclosure**: The methodology layer, GPU checklist, and book layer are all loaded on demand; only the minimal trigger and diagnostic logic stays resident, to save tokens.
+- **Do not create a pile of skills / rules right away**: The model already has many capabilities; use them fully first, then add structure.

package/references/agentic-workflow.md

@@ -1,3 +1,5 @@
+> **语言路由**：若用户消息为英文，请读取并遵循同目录下的 `agentic-workflow.en.md`。
+
 # Agentic 研究工作方式 / Human-in-the-Agent-Loop
 
 > 提炼自一线 auto-research 实践（Infra / 算法 / 芯片 Co-design）。它定义本 skill 期望的**协作思维方式**：人做什么、Agent 做什么、用什么循环把"激发数学能力"落地。

package/references/books/abstract-algebra.md

@@ -88,15 +88,15 @@
 | 有限域编码压缩 KV/梯度 | 22, 31 | 2❌ 非浮点 MAC，1/6✅ int 并行 | ❌主干 / ✅前后处理 |
 | 循环群表示（FFT/RoPE token mixing）| 4, 11 | 1✅ 2✅ 3✅（n log n）| ✅ |
 
-## 该调用哪个思想武器
+## 该调用哪个思想透镜
 
-- **主：`symmetry-invariance`（对称与不变性）**——群作用、等变/不变、轨道-稳定子是本书与 ML 的最大接口。
-- **副：`abstraction`**——提取群/环/域的公共结构，看穿"不同模块其实是同一代数对象"。
+- **主：`symmetry`（对称与不变性）**——群作用、等变/不变、轨道-稳定子是本书与 ML 的最大接口。
+- **副：`categorical`**——提取群/环/域的公共结构，看穿"不同模块其实是同一代数对象"。
 - **副：`axiomatization`**——放松环公理得半环、逐条核对所设代数性质是否真成立（防伪对称）。
-- **副：`transformation`**——同态/同构作等价转换、FFT/频域变换简化问题。
+- **副：`duality`**——同态/同构作等价转换、FFT/频域变换简化问题。
 - **副：`discrete-combinatorial`**——有限群计数（Burnside, 29）、编码（31）、有限域枚举（22）。
 
-典型组合链：先 `symmetry-invariance` 识别问题里的群与不变量 → `abstraction` 抽出公共代数结构 → `axiomatization` 核对公理是否真成立（防伪对称、防误用减法）→ `transformation` 落成可学习线性映射 → 最后过 `../gpu-friendly-math.md` 八维门。
+典型组合链：先 `symmetry` 识别问题里的群与不变量 → `categorical` 抽出公共代数结构 → `axiomatization` 核对公理是否真成立（防伪对称、防误用减法）→ `duality` 落成可学习线性映射 → 最后过 `../gpu-friendly-math.md` 八维门。
 
 ## 反模式
 
@@ -110,6 +110,10 @@
 
 ## 深挖入口
 
+> **📖 书目信息**：Joseph A. Gallian, *Contemporary Abstract Algebra*, 8th Edition, Brooks/Cole, Cengage Learning, 2013. ISBN 978-1-133-59971-5.
+>
+> **启用方式**：将 `Contemporary Abstract Algebra.pdf` 放入项目根目录的 `math_book/` 文件夹，Agent 即可自动搜索原文。PDF 不随 npm/git 分发（版权原因），需自行获取。
+
 > **全保真回查 = Agent 自动搜索本地 PDF**：`math_book/Contemporary Abstract Algebra.pdf`。需要精确定义/定理/例子时，让 Agent 按下列真实章号定向略读，不要凭记忆；若安装环境没有 `math_book/`，停在本蒸馏稿层。
 
 - **Ch 5 Permutation Groups**——轨道 orbit / 稳定子 stabilizer：等变权重共享的根。

package/references/books/algebraic-geometry-rising-sea.md

@@ -113,7 +113,7 @@
 ## GPU 友好性警告
 
 > **必读且唯一权威**：`../gpu-friendly-math.md`
-> 八维记分卡：①张量化 ②GEMM 可映射 ③复杂度 ④显存/KV ⑤低精度稳定 ⑥并行通信 ⑦稀疏结构 ⑧算子融合。
+> 八维记分卡：①张量化 ②GEMM 可映射 ③复杂度 ④显存/KV ⑤低精度稳定 ⑥并行与通信 ⑦稀疏结构 ⑧算子融合。
 > **数学美 ≠ 可算**；任一维"不友好且不可改造"即淘汰。
 
 **可落 GEMM / 亚二次（友好 ✅）：**
@@ -135,12 +135,12 @@
 - **理想消元 / Gröbner 基（Chevalley 定理与消元理论，§8.4）**：组合爆炸、串行、非结构化 → 违反维度③⑥⑦。
 - **动态 / 非结构化的 Čech 覆盖**：若覆盖随输入变化或交叠不规则，H¹ 计算退化为随机 gather/scatter → 违反维度⑦。**必须固定有限覆盖 + 结构化（块/带状）交叠。**
 
-## 该调用哪个思想武器
+## 该调用哪个思想透镜
 
 - **topological-thinking（🌀 拓扑思想）** — 主力：上同调、H¹ 障碍、连续变形下的不变量，正是层/上同调激活的母题。
-- **abstraction（🧩 抽象化）**：用"层 = 局部数据 + 黏合"提取消息传递的本质，把工程算子抽象为限制映射。
-- **transformation（🔄 变换思想）**：tropicalization、Plücker 嵌入、pullback/pushforward —— 用等价转换把难算问题搬到可算坐标系。
-- **symmetry-invariance（⚛️ 对称与不变性）**：射影不变、规范（gauge）对称、层的协变性，约束模型的等变结构。
+- **categorical（范畴化透镜）**：用"层 = 局部数据 + 黏合"提取消息传递的本质，把工程算子抽象为限制映射。
+- **duality（对偶透镜）**：tropicalization、Plücker 嵌入、pullback/pushforward —— 用等价转换把难算问题搬到可算坐标系。
+- **symmetry（⚛️ 对称与不变性）**：射影不变、规范（gauge）对称、层的协变性，约束模型的等变结构。
 - **axiomatization（📐 公理化）**：把层公理（局部性 + 黏合）、上同调长正合列当作正确性约束，审查"该有的一致性是否被违反"。
 
 ## 反模式
@@ -154,6 +154,10 @@
 
 ## 深挖入口
 
+> **📖 书目信息**：Ravi Vakil, *The Rising Sea: Foundations of Algebraic Geometry*, Princeton University Press, 2025. ISBN 978-0-691-26866-8.
+>
+> **启用方式**：将 `The Rising Sea Foundations of Algebraic Geometry.pdf` 放入项目根目录的 `math_book/` 文件夹，Agent 即可自动搜索原文。PDF 不随 npm/git 分发（版权原因），需自行获取。
+
 **全保真回查（full-fidelity lookup）**：当摘要不足以支撑设计时，让 Agent 自动检索本地 PDF
 `math_book/The Rising Sea Foundations of Algebraic Geometry.pdf`
 （用 `pdftotext -f <start> -l <end>` 取定向页，**不要 dump 整本**）。

package/references/books/differential-geometry.md

@@ -98,15 +98,15 @@
 
 **结论**：黎曼 / 信息几何方法**只有在度量被结构化因子化（Kronecker / 块对角 / 低秩）后才 GPU 可行**；精确求逆与精确平行移动都属于"美但不可算"，须改造或淘汰。
 
-## 该调用哪个思想武器
+## 该调用哪个思想透镜
 
-- **symmetry-invariance（⚛️ 对称与不变性）—— 首选。** 规范等变、李群对称、纤维丛 = 把"frame / 坐标选择无关"编码成对称性，是本书与 DL 最强的接口。
+- **symmetry（⚛️ 对称与不变性）—— 首选。** 规范等变、李群对称、纤维丛 = 把"frame / 坐标选择无关"编码成对称性，是本书与 DL 最强的接口。
 - **optimization（⚖️ 优化）—— 并列首选。** 自然梯度、Riemannian SGD、曲率正则都是"在弯曲约束空间里找最优"。
 - **transformation（🔄 变换）**：指数 / 对数映射、retraction、坐标变换简化问题。
 - **modeling（🌉 建模）**：把参数 / 数据空间显式建模为流形，再翻译回算法。
 - **topological-thinking（🌀 拓扑）**：辅助——de Rham 上同调 / 整体不变量用于守恒量与可积性诊断。
 
-组合建议：先 `symmetry-invariance` 定对称结构 → `optimization` 落到自然梯度 / Riemannian 优化 → `transformation` 处理 retraction → 过 `../gpu-friendly-math.md` 验收门。
+组合建议：先 `symmetry` 定对称结构 → `optimization` 落到自然梯度 / Riemannian 优化 → `duality` 处理 retraction → 过 `../gpu-friendly-math.md` 验收门。
 
 ## 反模式
 
@@ -127,6 +127,10 @@
 
 ## 深挖入口
 
+> **📖 书目信息**：Jeffrey M. Lee, *Manifolds and Differential Geometry*, Graduate Studies in Mathematics Vol. 107, American Mathematical Society, 2009. ISBN 978-0-8218-4815-9.
+>
+> **启用方式**：将 `Manifolds and Differential Geometry.pdf` 放入项目根目录的 `math_book/` 文件夹，Agent 即可自动搜索原文。PDF 不随 npm/git 分发（版权原因），需自行获取。
+
 > 全保真回查 = 让 Agent **自动搜索本地 PDF** `math_book/Manifolds and Differential Geometry.pdf`（按章号 / 节号定位，勿凭记忆复述）。本摘要只给坐标，不替代原文。
 
 - **§6.8 Principal and Associated Bundles** + **§12.12 G-Connections**：规范等变的几何基础（主丛 + 联络 = 规范场）。