[RNE Rewrite] refactor!: add better model schema contract and validation logic

.agents/skills/README.md

@@ -7,6 +7,6 @@ This directory contains specialized skills (recipes) to guide contributors and A
 - [Core Guidelines](./core-guidelines/SKILL.md) — Symmetrical architecture, conventions, and model rules.
 - [Add Native Extension](./add-native-extension/SKILL.md) — C++ operations and JSI bindings.
 - [Add Task Pipeline](./add-task-pipeline/SKILL.md) — TypeScript task pipelines and React hooks.
-- [Model Schema Validation](./model-schema-validation/SKILL.md) — SymbolicTensor schemas and validation.
+- [Model Schema Validation](./model-schema-validation/SKILL.md) — Model specs, dynamic shapes, and schema validation.
 - [Verify and Build](./verify-and-build/SKILL.md) — TypeScript typechecking, native rebuilding, and troubleshooting.
 - [Skills Maintenance](./skills-maintenance/SKILL.md) — Keeping skills synchronized with core primitives.

.agents/skills/add-task-pipeline/SKILL.md

@@ -59,17 +59,17 @@ When implementing task constructors like `create<Task>` (e.g. `createClassifier`
    - Handle model-specific configuration parameters (such as unique normalization factors, thresholds, or label arrays) dynamically through the TypeScript task options argument rather than baking them rigidly into JSI C++ code or the model structure. This rule contrasts TypeScript options against values baked into C++ or the model; to choose between a TypeScript **option** and a TypeScript **constant**, see Principle 6.
 
 6. **Options vs. Constants (bucket by who varies the value)**:
-   - Every parameter belongs in exactly one of three places. Decide by asking *who varies this, and when*:
+   - Every parameter belongs in exactly one of three places. Decide by asking _who varies this, and when_:
 
-     | The value...                                                                         | Lives as                                                       | Example                                                        |
-     | ------------------------------------------------------------------------------------ | -------------------------------------------------------------- | -------------------------------------------------------------- |
-     | Varies across the shipped models/variants of the pipeline                             | A task **option**, set per-model in `models.ts`                 | Whisper `tiny`/`base`/`small` sizes; normalization factors; labels |
-     | Is fixed by the model architecture or the `.pte` export contract, identical for every shipped variant | A **`const`** in the task file, beside the code that reads it   | Static tensor shapes; export-pinned scheduler/decoder scalars   |
-     | Is a per-call choice made by the app developer                                        | An **argument** to the worklet executor                         | `threshold`, `seed`, `prompt`                                   |
+     | The value...                                                                                          | Lives as                                                      | Example                                                            |
+     | ----------------------------------------------------------------------------------------------------- | ------------------------------------------------------------- | ------------------------------------------------------------------ |
+     | Varies across the shipped models/variants of the pipeline                                             | A task **option**, set per-model in `models.ts`               | Whisper `tiny`/`base`/`small` sizes; normalization factors; labels |
+     | Is fixed by the model architecture or the `.pte` export contract, identical for every shipped variant | A **`const`** in the task file, beside the code that reads it | Static tensor shapes; export-pinned scheduler/decoder scalars      |
+     | Is a per-call choice made by the app developer                                                        | An **argument** to the worklet executor                       | `threshold`, `seed`, `prompt`                                      |
 
-   - **A parameter with exactly one valid value is not an option — it is a constant.** For single-model pipelines (e.g. `sdxsTextToImage`), exposing export-pinned scalars or static shapes as options advertises knobs that either fail schema validation or silently corrupt output when touched. Prefer a `const` with a comment naming *why* the value is fixed.
-   - **Corollary (a quick smell test):** if every variant in `models.ts` passes an *identical* options object, those fields are not configuration — move them into the task file as constants and shrink the model type to the paths that actually differ.
-   - Do not keep a loop, parameter, or code path solely because it *looks* more general. If the surrounding math is only valid for one value (e.g. coefficients pinned to a single distilled timestep), the generality is fake and the parameter is a correctness trap.
+   - **A parameter with exactly one valid value is not an option — it is a constant.** For single-model pipelines (e.g. `sdxsTextToImage`), exposing export-pinned scalars or static shapes as options advertises knobs that either fail schema validation or silently corrupt output when touched. Prefer a `const` with a comment naming _why_ the value is fixed.
+   - **Corollary (a quick smell test):** if every variant in `models.ts` passes an _identical_ options object, those fields are not configuration — move them into the task file as constants and shrink the model type to the paths that actually differ.
+   - Do not keep a loop, parameter, or code path solely because it _looks_ more general. If the surrounding math is only valid for one value (e.g. coefficients pinned to a single distilled timestep), the generality is fake and the parameter is a correctness trap.
 
 ## 🚫 Avoid / Anti-Patterns
 
@@ -95,7 +95,7 @@ import type { WorkletRuntime } from 'react-native-worklets';
 
 import { tensor } from '../../../core/tensor';
 import { loadModel } from '../../../core/model';
-import { validateModelSchema, SymbolicTensor } from '../../../core/modelSchema';
+import { validateSpec, method, f32 } from '../../../core/schema';
 import { wrapAsync } from '../../../core/runtime';
 import { type ImageBuffer } from '../image';
 import { createImagePreprocessor, type ImagePreprocessorOptions } from './preprocessing';
@@ -137,15 +137,15 @@ export async function createMyTask(
   const { modelPath, taskOpts } = config;
   const model = await wrapAsync(loadModel, runtime)(modelPath);
 
-  // Validate model schema
-  const meta = validateModelSchema(
-    model,
-    'forward',
-    [SymbolicTensor('float32', [1, 3, 'H', 'W'], [3, 'H', 'W'])],
-    [SymbolicTensor('float32', [1, 10], [10])]
-  );
-  const inpShape = meta.inputTensorMeta[0]!.shape;
-  const outShape = meta.outputTensorMeta[0]!.shape;
+  // Validate model spec
+  const { variant, dims } = validateSpec(model.schema, {
+    batched: method('forward', [f32(1, 3, 'H', 'W')], [f32(1, 10)]),
+    unbatched: method('forward', [f32(3, 'H', 'W')], [f32(10)]),
+  });
+
+  const [H, W] = dims.constant('H', 'W');
+  const inpShape = { batched: [1, 3, H, W], unbatched: [3, H, W] }[variant];
+  const outShape = { batched: [1, 10], unbatched: [10] }[variant];
 
   // 2. Pre-allocate static tensors
   const tensors = [tensor('float32', outShape)] as const;

.agents/skills/core-guidelines/SKILL.md

@@ -72,7 +72,7 @@ Use the following index to locate the specific procedural guides for your task:
 | **Add a new native operator or C++ binding** | [SKILL.md](../add-native-extension/SKILL.md)    | Procedural guide to implementing C++ functions, exposing them via JSI, and writing TypeScript bridge wrappers. |
 | **Create a task pipeline or hook**           | [SKILL.md](../add-task-pipeline/SKILL.md)       | Guide to building end-to-end TS pipelines (e.g. object detection) and exposing them via React hooks.           |
 | **Verify, rebuild, or troubleshoot changes** | [SKILL.md](../verify-and-build/SKILL.md)        | Workflows for rebuilding TS/C++ and resolving common JSI runtime errors.                                       |
-| **Validate model constraints & schemas**     | [SKILL.md](../model-schema-validation/SKILL.md) | Guide on specifying SymbolicTensor constraints and shapes for model signature validation.                      |
+| **Validate model constraints & schemas**     | [SKILL.md](../model-schema-validation/SKILL.md) | Guide on specifying model specs, dynamic shapes, and runtime constraints for model validation.                 |
 | **Maintain or refactor codebase patterns**   | [SKILL.md](../skills-maintenance/SKILL.md)      | Guide to keeping workspace skills in sync with codebase state to prevent documentation decay.                  |
 
 ---