diff --git a/TODO_SOLID.md b/TODO_SOLID.md
new file mode 100644
index 0000000..92cedb8
--- /dev/null
+++ b/TODO_SOLID.md
@@ -0,0 +1,114 @@
+# SOLID Principles Analysis & Refactoring TODO
+
+This document outlines the identified violations of SOLID principles within the project's architecture. The issues are prioritized by severity to guide refactoring efforts.
+
+---
+
+## Part 1: CSS Processing Architecture Analysis
+
+### High Severity
+
+#### 1. Violation: Interface Segregation Principle (ISP) & Single Responsibility Principle (SRP)
+
+- **Component:** `StyleAccumulator` interface and `ComputedStyle` class (`src/css/style.ts`).
+- **Problem:** The `StyleAccumulator` is a "fat interface" that includes dozens of optional properties for every possible CSS style. Consequently, all `PropertyParser` implementations are forced to depend on this massive interface. The `ComputedStyle` class becomes a "God Object" responsible for managing all style-related data, leading to very low cohesion.
+- **Impact:** High coupling, poor maintainability, and difficult testing.
+- **Suggested Refactoring:** Decompose `StyleAccumulator` into smaller, cohesive interfaces (`FontStyles`, `BoxModelStyles`, etc.) and refine parser signatures to depend only on the interfaces they need.
+
+---
+
+### Medium Severity
+
+#### 2. Violation: Open/Closed Principle (OCP) & Single Responsibility Principle (SRP)
+
+- **Component:** `registerAllPropertyParsers` function (`src/css/parsers/register-parsers.ts`).
+- **Problem:** This is a centralized function that must be modified to add any new CSS property parser. This violates OCP (not closed for modification) and SRP (knows about every parser).
+- **Impact:** Extensibility bottleneck and reduced modularity.
+- **Suggested Refactoring:** Implement an auto-registration mechanism where parser modules can register themselves, removing the need for a central registration function.
+
+---
+---
+
+## Part 2: PDF Generation Architecture Analysis
+
+### High Severity
+
+#### 1. Violation: Dependency Inversion Principle (DIP) & Open/Closed Principle (OCP)
+
+- **Component:** `PagePainter` class (`src/pdf/page-painter.ts`).
+- **Problem:** High-level modules like `PagePainter` depend directly on concrete low-level implementations (e.g., `TextRenderer`), instantiating them directly. This violates DIP and OCP, as new renderers cannot be added without modifying `PagePainter`.
+- **Impact:** High rigidity, strong coupling, and untestable code.
+- **Suggested Refactoring:** Introduce renderer interfaces (e.g., `ITextRenderer`) and use Dependency Injection (DI) to provide concrete implementations to the `PagePainter` constructor.
+
+#### 2. Violation: Single Responsibility Principle (SRP)
+
+- **Component:** `convertNode` function (`src/pdf/layout-tree-builder.ts`).
+- **Problem:** This is a "God function" that centralizes the conversion logic for all types of nodes and styles (text, images, backgrounds, shadows, etc.).
+- **Impact:** Low cohesion, high complexity, and fragility.
+- **Suggested Refactoring:** Decompose the function by responsibility or use a Strategy Pattern where different "Node Converters" can be registered for different node types.
+
+---
+
+### Medium Severity
+
+#### 3. Violation: Interface Segregation Principle (ISP)
+
+- **Component:** `RenderBox` interface (`src/pdf/types.ts`).
+- **Problem:** `RenderBox` is a "fat interface" aggregating properties for all possible node types, forcing components to depend on irrelevant data.
+- **Impact:** Unnecessary coupling and reduced type safety.
+- **Suggested Refactoring:** Refactor `RenderBox` into a discriminated union type (`type RenderNode = ContainerNode | TextNode | ImageNode;`) where each type is specialized.
+
+#### 4. Violation: Single Responsibility Principle (SRP)
+
+- **Component:** `renderPdf` function (`src/pdf/render.ts`).
+- **Problem:** The function orchestrates the rendering pipeline but also handles setup logic (font initialization, header/footer layout).
+- **Impact:** Reduced cohesion, making the code harder to follow and test.
+- **Suggested Refactoring:** Separate configuration from orchestration. Move setup logic into a "context" object that is passed to `renderPdf`.
+
+---
+---
+
+## Part 3: Layout Architecture Analysis
+
+### No Violations Found
+
+- **Conclusion:** The layout architecture, centered around the `LayoutEngine` and the `LayoutStrategy` interface (`src/layout/pipeline/`), is an exemplary model of SOLID design within this project.
+- **Key Strengths:**
+    - **SRP:** `LayoutEngine` orchestrates, while individual strategies handle specific layout algorithms.
+    - **OCP & DIP:** The engine depends on the `LayoutStrategy` abstraction, and concrete strategies are provided via Dependency Injection. This makes the system fully extensible to new layout types without modifying the engine.
+    - **LSP & ISP:** The `LayoutStrategy` interface is a simple, robust contract that is well-defined and correctly implemented by all strategies.
+- **Recommendation:** This architecture should be used as a reference or blueprint for refactoring the CSS and PDF modules. The patterns used here (Strategy Pattern, Dependency Injection) would directly solve many of the issues identified in the other modules.
+
+---
+---
+
+## Part 4: Font Architecture Analysis (`src/pdf/font`)
+
+### High Severity
+
+#### 1. Violation: Open/Closed Principle (OCP) & Dependency Inversion Principle (DIP)
+
+- **Component:** The entire font embedding architecture, particularly `FontEmbedder` (`embedder.ts`) and `FontRegistry` (`font-registry.ts`).
+- **Problem:** The architecture is rigidly designed for the TTF font format. High-level modules like `FontRegistry` directly instantiate and depend on the concrete low-level `FontEmbedder`, which itself contains TTF-specific logic. There are no abstractions for font parsers or embedders.
+- **Impact:**
+    - **Rigidity:** The system is not open to extension with new font formats (like WOFF2 or OTF) without significant modifications to existing code.
+    - **High Coupling & Untestable Code:** `FontRegistry` is tightly coupled to `FontEmbedder`, making it very difficult to unit test in isolation.
+- **Suggested Refactoring:**
+    1.  **Introduce Abstractions:** Create interfaces like `IFontParser` and `IFontEmbedder`.
+    2.  **Implement Strategy Pattern:** Create concrete classes like `TtfParser` and `TtfEmbedder` that implement these interfaces.
+    3.  **Use Dependency Injection:** Refactor `FontRegistry` to depend on the `IFontEmbedder` interface, and inject the concrete implementation.
+
+---
+
+### Medium Severity
+
+#### 2. Violation: Single Responsibility Principle (SRP)
+
+- **Component:** `FontRegistry` class and `FontEmbedder` class.
+- **Problem:**
+    - `FontRegistry`: Accumulates multiple responsibilities: font caching, font name resolution/aliasing, font variant selection, and orchestration of embedding.
+    - `FontEmbedder`: Its `embedFont` method is a "God method" that handles the low-level construction of multiple complex PDF objects.
+- **Impact:** The classes have low cohesion, making them harder to understand, maintain, and test.
+- **Suggested Refactoring:**
+    - **For `FontRegistry`:** Extract the logic for font name resolution and variant selection into a separate `FontResolver` class.
+    - **For `FontEmbedder`:** Decompose the `embedFont` method by extracting the logic for creating PDF objects into dedicated builder or factory functions.