qualcomm

Rebase the current branch into clean conventional commits with proper author attribution. Use when the user wants to clean up a branch's commit history into well-organized conventional commits.

Use when writing, reviewing, or refactoring Angular component or directive tests, especially tests for inputs, outputs, host behavior, template probes, forms, accessibility, or signal-backed behavior.

Use when writing, reviewing, or refactoring Angular tests for application behavior, services, routes, state, forms, pipes, or integration flows, especially when tests may assert implementation details instead of public outcomes.

Use when writing, reviewing, or refactoring React component tests, especially tests for props, callbacks, controlled state, children, accessibility, form behavior, or multiple public component API variants.

Use when writing, reviewing, or refactoring React tests for application behavior, hooks, routes, state, or integration flows, especially when tests may assert implementation details instead of public outcomes.

Use when writing, reviewing, or refactoring tests that run in Vitest Browser Mode, especially tests using page, userEvent, expect.element, browser locators, polling assertions, or real DOM interactions.

Use when writing or modifying QUI Angular template usage involving q-icon-button, q-inline-icon-button, q-header-bar-action-icon-button, q-avatar-image, labels for q-text-input/q-number-input/q-password-input/q-select/q-combobox/q-switch/q-checkbox/q-radio, interactive q-card, q-alert-banner actions, q-card actions, q-header-bar-action-bar, q-menu-trigger, or q-select-root checkbox indicators.

Use when reviewing or final-checking QUI Angular template usage involving q-icon-button, q-inline-icon-button, q-header-bar-action-icon-button, q-avatar-image, labels for q-text-input/q-number-input/q-password-input/q-select/q-combobox/q-switch/q-checkbox/q-radio, interactive q-card, q-alert-banner actions, q-card actions, q-header-bar-action-bar, q-menu-trigger, or q-select-root checkbox indicators.

Fallback skill for porting x64 code to ARM64 when no specific spec matches. Provides mandatory architectural constraints that ALL ARM64 migrations must respect regardless of code pattern — covering both hand-written assembly and C/intrinsics output. Use when the dispatcher receives a porting_item with port_spec_ids: [] and mode=llm-freeform, OR when the LLM is producing freeform ARM64 output (C, intrinsics, or asm) without a more specific leaf skill. Ensures freeform porting still adheres to ARM64 correctness invariants: Windows ARM64 ABI, weak memory ordering, NEON 128-bit width limits, ARM64EC ABI shims, and intrinsic header / target-macro hygiene.

Convert ARM64 NEON inline assembly written in Clang/GCC-style `asm volatile(...)` blocks into maintainable C/C++ intrinsics, and must build a Verification project that compares the original asm against the translated intrinsics for semantic equivalence. Use when: (1) the user provides a source file path and line range and wants a specific ARM64 asm block rewritten as intrinsics; (2) the main source file needs an MSVC-compatible `#if defined(_MSC_VER) && (defined(_M_ARM64) || defined(_M_ARM64EC))` intrinsics branch while preserving the original asm in `#else`; (3) the task requires a mandatory GoogleTest-based verification workflow using the bundled template stored under `assets/Verification/`. Keywords: asm to intrinsics, ARM64 asm to intrinsics, NEON intrinsics, clang asm, msvc arm64, verification, gtest equivalence.

Generate an x64-to-ARM64 porting report in YAML format. Triggers when the user asks to generate/create/write an ARM64 porting report or migration analysis. Trigger phrases include: generate arm64 porting report, create porting report, write porting report, arm64 porting analysis, generate migration report, porting analysis for this code.

Convert x64 assembly to AArch64 arm64 — both inline asm (asm volatile / __asm__) and standalone MASM .asm files (PROC/ENDP, equ, macro/endm) translated to GAS .S form. Use when porting x64-specific asm to arm64, adding arm64 support alongside an existing x64 path, translating an entire MASM hot-loop file, or auditing code for missing arm64 coverage.

Expert skilltree layer-1 dispatcher. Receives porting task commands from tasks.md, parses arguments, loads porting_item data and spec descriptions, then routes to the appropriate leaf skill for actual migration execution. Invoked as: /dispatcher-skill <porting_item_id> --specs <ids> --source <path> or /dispatcher-skill <porting_item_id> --mode llm-freeform --source <path>. Does NOT contain migration logic itself — only routing and data assembly.

Bridges EasyWoS scan output with OpenSpec task generation for x64-to-ARM64 porting. MUST be invoked when OpenSpec generates tasks.md for changes involving porting_items. Parses EasyWoS YAML, runs hybrid spec matching (script initial screening + LLM semantic refinement), produces a matched YAML file with port_spec_ids and confidence scores, then guides tasks.md generation with dispatcher skill invocations. On invocation it FIRST runs a dependency preflight check and stops to ask the user if any dependent skill or data file is missing. Triggered via openspec/config.yaml rules.tasks.

Guide for enabling Windows ARM64 support in software projects. This skill will be used when users want to "enable ARM64", "add ARM64 support", "support Windows on ARM", or "make project ARM64 compatible". It detects build systems (cmake, visual_studio, autotools, bazel, make, ninja, qmake, scons) with priority ordering, checks if ARM64 is already supported, and modifies build files to add ARM64 configurations.

Expert guidance for porting x86/x64 SSE/AVX vectorized C++ to ARM64 NEON, grounded in the Microsoft STL's two real backends (vector_algorithms.cpp). Use this skill when: translating __m128i/__m256i intrinsics to NEON (uint8x16_t/uint16x8_t/etc.), replacing _mm_*/_mm256_* with vld1q/vst1q/vceqq/vmaxvq/vminq, converting AVX2 tail-mask patterns (_Avx2_tail_mask_32) to ARM64 descending-granularity loops, replacing __isa_enabled with IsProcessorFeaturePresent, removing _Zeroupper_on_exit, handling 64-bit int min/max exclusion on ARM64, fixing _M_ARM64-only guards that miss _M_ARM64EC, auditing #ifdef guards for cross-arch correctness, or any x86-to-ARM64 SIMD migration in Windows C++ codebases. Also trigger for NEON equivalents of SSE/AVX ops, or why an x64 pattern cannot port to ARM64.

Add a new LLM model to the geniex runtime (creates spec header, example executable, CMakeLists)

Reference for developing against the QAIRT/QNN runtime — graph execution, KV cache, tensor I/O, multi-shard wiring

Convert xtensor prototype code to production QNN direct-buffer operations (zero-copy)

Write C++ tensor logic using xtensor (NumPy-like API) for prototyping before QNN conversion

GenieAPIService technical documentation retrieval. Find guides on platform deployment, model configuration, and API usage. GenieAPIService is an OpenAI-compatible API service that enables running large language models(include LLM & VLM model) locally on Qualcomm(高通) Windows on Snapdragon, Android, and Linux platforms. It leverages the device local NPU(HTP) or CPU for efficient inference.

QAI AppBuilder technical documentation retrieval. Find guides on installation, Python/C++ APIs, and model deployment examples. QAI AppBuilder is a rapid AI application development framework designed to simplify the deployment of QNN models on NPU (HTP) across Qualcomm(高通) Windows on Snapdragon, Android, and Linux platforms. This tool is highly suitable for deploying classic models (all types of models except large language models can be deployed via QAI AppBuilder), such as real_esrgan_x4plus, inception_v3, beit, easy_ocr, and whisper_base_en. This tool is only applicable for loading QNN (*.bin) format models and performing inference, and is not suitable for converting model formats.

AIPC, AI Porting Conversion. Tools and workflows for QAIRT/AIPC project setup, model conversion, inspection, operator patching, quantization, context-binary generation, and inference on Qualcomm platforms. Use this skill when creating or initializing an AIPC project, exporting AI models to ONNX, converting ONNX models to QNN or SNPE/DLC, converting models to FP16/FP32, patching unsupported operators, generating context binaries, or implementing inference for QNN/SNPE DLC.

Generate an HTML coverage report for all Go tests via bazel + genhtml, then open it in the browser.

Reorganize the commits on the current PR branch into a clean, logically-grouped history with short one-line messages. Use when the user asks to tidy, squash, reshape, or clean up commits on the current branch or PR. ALSO proactively suggest this skill (ask, do not run unprompted) when the user signals a PR is finished or ready to review — e.g. "done with this PR", "ready for review", "let's open the PR", or just before calling `gh pr ready` / marking a draft ready — especially if `git log <base>..HEAD --oneline` shows many small WIP / fixup / "oops" commits.

Guide for enabling Windows ARM64 support in software projects. This skill will be used when users want to "enable ARM64", "add ARM64 support", "support Windows on ARM", or "make project ARM64 compatible". It detects build systems (cmake, visual_studio, autotools, bazel, make, ninja, qmake, scons) with priority ordering, checks if ARM64 is already supported, and modifies build files to add ARM64 configurations.