Add maintainer docs and publish handoff guidance

Author: leehack

AGENTS.md

@@ -0,0 +1,54 @@
+# AGENTS.md
+
+Guidance for coding agents working in `llama-web-bridge`.
+
+## Scope and Ownership
+
+- This repository owns WebGPU bridge source/runtime build for llama.cpp web.
+- It publishes versioned assets to `llama-web-bridge-assets` via workflow.
+- `llamadart` consumes those published assets.
+
+## Related Repositories
+
+Common maintainer sibling layout:
+
+```text
+../llamadart
+../llamadart-native
+../llama-web-bridge
+../llama-web-bridge-assets
+```
+
+## Build Commands
+
+```bash
+./scripts/build_bridge.sh
+```
+
+Useful environment overrides:
+
+- `LLAMA_CPP_DIR`
+- `BUILD_DIR`
+- `OUT_DIR`
+- `CMAKE_BUILD_TYPE`
+
+## CI / Release
+
+- CI build gate: `.github/workflows/ci.yml`
+- Publish workflow: `.github/workflows/publish_assets.yml`
+  - Requires `WEBGPU_BRIDGE_ASSETS_PAT`
+  - Pushes assets + tag to `llama-web-bridge-assets`
+
+## Change Boundaries
+
+- Keep runtime bridge code in `js/` and `src/`.
+- Keep publishing logic in workflow only.
+- Do not edit assets repository files from here outside publish flow.
+
+## Cross-Repo Handoff to `llamadart`
+
+After publishing assets tag:
+
+1. Update/fetch pinned bridge assets in `llamadart`:
+   `WEBGPU_BRIDGE_ASSETS_TAG=<tag> ./scripts/fetch_webgpu_bridge_assets.sh`
+2. Update docs/changelog in `llamadart` if behavior changed.

CONTRIBUTING.md

@@ -0,0 +1,54 @@
+# Contributing to llama-web-bridge
+
+Thanks for contributing.
+
+## Purpose
+
+This repo builds the reusable JS/WASM bridge runtime for llama.cpp web usage.
+Published artifacts are consumed from `llama-web-bridge-assets`.
+
+## Prerequisites
+
+- Emscripten SDK (`emcmake`, `emcc`)
+- CMake toolchain
+- Access to a llama.cpp checkout
+
+## Setup
+
+```bash
+git clone https://github.com/leehack/llama-web-bridge.git
+cd llama-web-bridge
+./scripts/build_bridge.sh --help
+```
+
+## Local Build
+
+```bash
+./scripts/build_bridge.sh
+# or
+LLAMA_CPP_DIR=../llama.cpp OUT_DIR=dist ./scripts/build_bridge.sh
+```
+
+## Validate Outputs
+
+Expected files:
+
+- `dist/llama_webgpu_bridge.js`
+- `dist/llama_webgpu_core.js`
+- `dist/llama_webgpu_core.wasm`
+
+## Publish Process
+
+Use workflow `.github/workflows/publish_assets.yml`:
+
+1. Set input `assets_tag` (new tag).
+2. Optionally set `assets_repo` and `llama_cpp_tag`.
+3. Ensure `WEBGPU_BRIDGE_ASSETS_PAT` secret is configured.
+4. Workflow builds, generates `manifest.json`/`sha256sums.txt`, pushes to
+   assets repo, and creates matching tag there.
+
+## Repository Boundaries
+
+- Bridge runtime source/build belongs here.
+- Versioned static artifacts belong in `llama-web-bridge-assets`.
+- Consumer integration (loading/fallback behavior) belongs in `llamadart`.

README.md

@@ -69,3 +69,8 @@ After publish, assets are CDN-available at:
 - `https://cdn.jsdelivr.net/gh/leehack/llama-web-bridge-assets@v0.1.1/llama_webgpu_bridge.js`
 - `https://cdn.jsdelivr.net/gh/leehack/llama-web-bridge-assets@v0.1.1/llama_webgpu_core.js`
 - `https://cdn.jsdelivr.net/gh/leehack/llama-web-bridge-assets@v0.1.1/llama_webgpu_core.wasm`
+
+## Maintainer Docs
+
+- `AGENTS.md`: agent workflow and cross-repo handoff
+- `CONTRIBUTING.md`: contributor setup/build/publish steps