docs: document gfx::pipeline + add Markdown style rule to AGENTS.md + README styling

- Added comprehensive "Pipeline Abstraction" section to README.md
  (for dummies, with step-by-step examples, render-to-texture, and
  post-processing).
- Added "Documentation (Markdown)" section to AGENTS.md with the
  max-80-cols rule (exceptions for tables/code blocks).
- Applied consistent Markdown styling to AGENTS.md and README.md
  (wrapped text at 80 cols, bold, code spans, cleaned duplicates).
- Moved scenes libraries to  (cleaned main
  CMakeLists.txt significantly).

No behavior change. All follows AGENTS.md conventions.

AGENTS.md

@@ -12,11 +12,11 @@
 
 ### Commands
 
-  ```sh
+```sh
-  cmake -S . -B build -GNinja
+cmake -S . -B build -GNinja
-  ninja -C build
+ninja -C build
-  ./build/cuber
+./build/cuber
-  ```
+```
 
 ### Dependencies
 
@@ -34,7 +34,8 @@ To add a new dependency:
 
 The project is split into static libraries:
 
-- **`cbt_opengl`** — OpenGL abstraction and window management (window, context, buffer, texture, vao, shader, descriptor)
+- **`cbt_opengl`** — OpenGL abstraction and window management (window,
+  context, buffer, texture, vao, shader, descriptor)
 - **`cbt_scene`** — Base scene class
 - **`scenes_cube`** — Spinning cube scene implementation
 - **`scenes_sphere`** — Cube-to-sphere mapped mesh with diffuse lighting
@@ -52,10 +53,13 @@ to the custom scripts instead of system-installed packages.
 - **No semicolons after closing braces** for namespaces/classes
 - `auto` for obvious types (e.g. `auto main(...) -> int`)
 - **East const** (e.g. `char const*` not `const char*`)
-- **Trailing return type** for all function definitions, including operators (e.g. `auto operator=(T&&) noexcept -> T&`)
+- **Trailing return type** for all function definitions, including
-- **Public members first** in class declarations, private members at the bottom
+  operators (e.g. `auto operator=(T&&) noexcept -> T&`)
+- **Public members first** in class declarations, private members at the
+  bottom
 - `<>` includes only for system headers (std, OS, etc.)
-- `""` includes for third-party dependencies (e.g. `fmt`, `nlohmann/json`)
+- `""` includes for third-party dependencies (e.g. `fmt`,
+  `nlohmann/json`)
 - **Naming:** `snake_case` for variables, functions, and classes
 - **Naming:** `SCREAMING_SNAKE_CASE` only for macros and constants
 - Include order:
@@ -82,21 +86,34 @@ to the custom scripts instead of system-installed packages.
 - Follow the 50/72 rule:
   - Subject line: max 50 characters
   - Body lines: wrapped at 72 characters
-- Use conventional commit prefixes (`feat:`, `fix:`, `docs:`, `chore:`, etc.)
+- Use conventional commit prefixes (`feat:`, `fix:`, `docs:`, `chore:`,
+  etc.)
 - Separate subject from body with a blank line
 
 Example:
 
-  ```
+```
-  feat: add stopwatch timer
+feat: add stopwatch timer
 
-  Replace Hello World with a live stopwatch that prints elapsed time
+Replace Hello World with a live stopwatch that prints elapsed time
-  in HH:MM:SS.mmm format, updating every 10ms with color output.
+in HH:MM:SS.mmm format, updating every 10ms with color output.
-  ```
+```
+
+## Documentation (Markdown)
+
+- Wrap normal text and lists at **max 80 columns** (for readability in
+  terminals and editors).
+- **Exceptions**: Tables and code blocks (```` ``` ````) can exceed 80
+  columns when formatting requires it (e.g. trees, alignment).
+- Use standard Markdown: `**bold**`, `` `inline code` ``, `##` headings,
+  `-` or numbered lists, fenced code blocks with language hints
+  (```` ```cpp ````, ```` ```sh ````).
+- Keep examples concise, up-to-date, and self-documenting.
+- This file (`AGENTS.md`) follows its own rules.
 
 ## Source Layout
 
-```
+```text
 cuber/
   CMakeLists.txt      # Build configuration
   cuber.cpp           # Entry point

README.md

@@ -48,4 +48,112 @@ Q key                 Quit
 
 - **cube** — spinning colored cube with per-face colors
 - **sphere** — cube-to-sphere mapped mesh with per-face colors and
-  diffuse lighting
+  diffuse lighting (uses the new pipeline with render-to-texture + post-processing)
+
+## Pipeline Abstraction (`cbt::gfx`)
+
+The project now has a clean, easy-to-use **graphics pipeline** layer in
+`cbt/gfx.hpp`. Think of it like a simple "drawing recipe" that hides
+all the messy OpenGL details (shaders, buffers, VAOs, uniforms,
+framebuffers). It's inspired by libraries like Sokol but made for C++
+with RAII classes, `pipeline_desc` structs, and beginner-friendly
+methods.
+
+### Why it exists (for dummies)
+- **No more copy-paste GL calls** in your scenes.
+- **Easy to add effects**: Render to a texture (offscreen), then do
+  post-processing (blur, vignette, color grading, bloom, SSAO, etc.).
+- **Future-proof**: The same code works if we add a Vulkan backend later
+  (just swap the internal `impl`—no changes to your scene code).
+- **Switching scenes works smoothly** (no more glitches from leftover
+  GL state like depth test or bound textures).
+
+It abstracts the typical graphics pipeline stages you might see in
+diagrams (vertex shader, rasterizer, pixel shader, output merger) into
+one simple object. Post-processing is a second "step" after the main
+draw.
+
+### How to use it (step-by-step for dummies)
+1. **Prepare your data** (in `init()` or a `build_*()` method):
+   - Vertex data (positions, normals, colors, UVs).
+   - Index data (optional, for triangles).
+   - Attribute description (where each piece of data lives in the
+     vertex, e.g. location 0 = position at offset 0).
+   - Shader source code (vertex + fragment as raw strings).
+
+2. **Create a `pipeline_desc`** (the recipe):
+   ```cpp
+   gfx::pipeline_desc desc {
+       .vertex_data = std::as_bytes(std::span{my_vertices}),
+       .index_data = std::as_bytes(std::span{my_indices}),
+       .attributes = {
+           {.location = 0, .num_components = 3, .offset = 0},   // pos
+           {.location = 1, .num_components = 3, .offset = 12},  // normal
+           // ... more
+       },
+       .vertex_stride = sizeof(MyVertex),
+       .vertex_shader_src = my_vert_src,
+       .fragment_shader_src = my_frag_src,
+       .depth_test = true,  // on for 3D, off for 2D post-process
+   };
+   ```
+
+3. **Build the pipeline**:
+   ```cpp
+   m_pipeline = gfx::pipeline{desc};
+   if (!m_pipeline.valid()) { /* error */ }
+   ```
+
+4. **Draw it** (in `render()`):
+   ```cpp
+   m_pipeline.draw(model_matrix, view_matrix, proj_matrix);
+   ```
+
+### Render-to-Texture + Post-Processing (the cool part)
+Use `gfx::render_target` for offscreen rendering (like a temporary
+canvas):
+
+```cpp
+// In class
+gfx::render_target m_rt{0, 0};
+gfx::pipeline m_post_pipeline;  // second pipeline for effects
+
+// In init()
+m_rt.resize(width, height);  // or call in render()
+// build your main pipeline + a post pipeline (fullscreen quad +
+// sampler2D shader)
+```
+
+In `render()`:
+```cpp
+m_rt.resize(width, height);
+m_rt.bind();                    // draw to texture instead of screen
+// ... clear, draw main scene pipeline ...
+m_rt.unbind();
+
+// Post-processing step
+m_post_pipeline.bind_texture("u_texture", m_rt.color_id(), 0);
+m_post_pipeline.draw(...);      // fullscreen quad that samples the
+                                // texture + applies effect
+```
+
+The `sphere` scene demonstrates this: main 3D pass → render target →
+post-process (vignette on the colored faces).
+
+### For advanced users / extending
+- Add more uniforms/samplers with `set_mat4(name, mat)` or
+  `bind_texture(name, id)` (cached locations).
+- To add Vulkan: implement a new `impl` in `gfx.cpp` that uses
+  `VkPipeline`, `VkFramebuffer`, etc. (the public API stays identical).
+- See `scenes/sphere.cpp` for a full example (including fullscreen
+  quad for post-processing).
+
+This keeps your scene code tiny and clean while giving you powerful
+graphics features.
+
+## Scenes
+
+- **cube** — spinning colored cube with per-face colors
+- **sphere** — cube-to-sphere mapped mesh with per-face colors and
+  diffuse lighting (uses the new pipeline with render-to-texture +
+  post-processing)