gpuni
Cross-platform GPU kernel dialect — write once, run on NVIDIA (CUDA), AMD (HIP), and Intel/others (OpenCL 1.2). Use when: (1) writing portable GPU code across CUDA/HIP/OpenCL, (2) porting CUDA kernels to AMD or OpenCL, (3) creating/editing *.gu.cu files, (4) needing cross-vendor GPU compute, (5) using address-space qualifiers (__global/__local/__constant), (6) implementing portable atomics or shared memory, (7) debugging OpenCL render errors.
python-developer
Эксперт Python разработки. Используй для Python best practices, async, typing и ecosystem.
numpy-interop
Protocols for cross-library data exchange including DLPack, buffer interfaces, and __array_ufunc__ for overriding NumPy functions. Triggers: DLPack, interoperability, __array_interface__, __array_ufunc__, buffer protocol.
complex-tensor-handler
Handle complex-valued tensors in PyTorch for astronomical imaging applications. This skill should be used when working with Fourier transforms, phase/amplitude representations, and complex arithmetic in PRISM.
load-anomaly-guard
Detects unsafe spikes in training load and emits SafetyFlags with conservative adjustments.
type-hint-adder
Add comprehensive type hints to Python functions and methods, including PyTorch tensor types. This skill should be used when improving code quality through static type checking or when preparing code for mypy validation.
zosia-temporal-awareness
We-layer temporal processing for Zosia. Provides efficient time awareness, gap detection, and event tracking. Feeds felt experiences to the I-layer.
pytorch-quantization
Techniques for model size reduction and inference acceleration using INT8 quantization, including Post-Training Quantization (PTQ) and Quantization Aware Training (QAT). (quantization, int8, qat, fbgemm, qnnpack, ptq, dequantize)
quantization
Model quantization for efficient inference and training. Covers precision types (FP32, FP16, BF16, INT8, INT4), BitsAndBytes configuration, memory estimation, and performance tradeoffs.
formula-translator
Translate Civilization 5 Vox Populi C++ formulas to TypeScript. Use when asked to translate a formula, add a new mechanic calculation, convert C++ code from CvGameCoreDLL, implement game logic like production costs, gold costs, growth thresholds, combat calculations, or yield computations. Handles GD_INT_GET/GD_FLOAT_GET macros, integer math preservation, and modifier chains.
astralrenderer-architecture
Guidelines for navigating and extending the AstralRenderer engine architecture
numpy-structured
Structured and record arrays for C-interoperability, binary blob interpretation, and multi-field tabular data handling. Triggers: structured array, record array, compound dtype, multi-field index.
ffi-memory-management
Your approach to handling ffi memory management. Use this skill when working on files where ffi memory management comes into play.
rust-coder
Rust expert for rainze_core PyO3 module. Use when working on Rust code, performance-critical components, or Python-Rust FFI.
autofix-symbol-dep
Fixes linker 'undefined reference' errors by finding and adding the library dependency.
operators-edge
Runs the Operator's Edge loop (State, Proof, Gates + adaptation checks). Use for any coding task, debugging, refactor, or system change.
ringpre-dev-feature
Lightweight 4-gate pre-dev workflow for small features (<2 days)
mmd-device-quad-cortex
Guide for using the Neural DSP Quad Cortex device library in MMD files. Use when the user mentions Quad Cortex, QC, Neural DSP guitar processor, or needs help with preset loading, scene switching, expression control, or stomp automation for the Quad Cortex.
