菜单
Cross-compile for Android (x-android-aarch64 preset), optionally build tests/benchmarks, and outline adb push/run steps. Use when building for Android targets or diagnosing Android toolchain issues.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Add or revise a model configuration JSON under model_configuration_files/ and ensure it is exercised by CTest (test/CMakeLists.txt) for the relevant backend; optionally onboard a new model end-to-end (downloads/sha256 entries, backend prompt/template quirks, tests, README updates). Use when onboarding a new model, updating supported models, or fixing model-related load/inference/test failures.
Build and run the benchmarking tools in this repository (including llm-bench-cli) across supported backends, and triage benchmark build/runtime issues (shared libs placement, model paths, threads/tokens, JNI off). Use when changing benchmark code, adding metrics, comparing performance, or verifying benchmark binaries for a backend.
Configure, build, and run CTest for this repository using CMake presets (native, x-android-aarch64, x-linux-aarch64), including common cache flags (LLM_FRAMEWORK, BUILD_JNI_LIB, USE_KLEIDIAI, BUILD_BENCHMARK, CPU_ARCH) and failure triage steps. Use when asked to run tests, reproduce build/CI failures, or diagnose CMake/CTest/JNI issues for this project.
Debug failing LLM integration tests caused by model output drift, incorrect context/runtime parameters (contextSize, batchSize, threads), prompt/template mismatches, or backend/framework regressions. Use when tests fail and you need to see the model response, reproduce a single failing CTest, or trace issues into src/cpp/frameworks (llama.cpp, onnxruntime-genai, mnn, executorch).
Run fast “session start / doctor” checks for this repository (toolchain + wiring sanity, framework version report, optional upstream update check), optionally generate a debug bundle, and when needed bump pinned backend framework versions with build+ctest verification. Use at session start or when upgrading llama.cpp/onnxruntime-genai/mnn/executorch pins.
Scaffold and integrate a new LLM backend/framework into this repository (new subdir under src/cpp/frameworks, wire LLM_FRAMEWORK routing, configuration options, downloads/models, tests, and docs). Use when adding a new framework, porting a backend, or refactoring backend glue code.
| name | llm-android-cross-build |
| description | Cross-compile for Android (x-android-aarch64 preset), optionally build tests/benchmarks, and outline adb push/run steps. Use when building for Android targets or diagnosing Android toolchain issues. |
Use this to build for Android targets using the x-android-aarch64 preset.
Windows note: if python3 isn’t available, use python (or py -3) for any scripts below.
During configure, CMake may use FetchContent to download dependencies from the network. If your network is blocked, configure can fail before tests run. To enable network access when the process runs in a sandbox, see skills/README.md. Otherwise, run the build on your local machine (with network access) and share the resulting build path with the tool (for example: build dir: /path/to/build), or point CMake at pre-downloaded dependencies in your environment.
cmake --preset=x-android-aarch64 -B build
If configure fails because the NDK is missing, set it explicitly:
export NDK_PATH=/path/to/ndk
cmake --preset=x-android-aarch64 -B build
Note: configure triggers downloads via the standard downloader; set HF_TOKEN (or ~/.netrc) for gated models.
Optional flags:
cmake --preset=x-android-aarch64 -B build -DLLM_FRAMEWORK=llama.cpp -DBUILD_JNI_LIB=ON
cmake --preset=x-android-aarch64 -B build -DBUILD_BENCHMARK=ON
cmake --build ./build --parallel
Use adb to copy binaries and models/configs to the device, then run them from a shell. Keep paths consistent with your config modelRoot.
NDK_HOME).JAVA_HOME and JDK version.TROUBLESHOOTING.md for platform-specific notes.