Loading...
Loading...
Found 108 Skills
· Configure/tune/migrate PostgreSQL, MongoDB, MySQL/MariaDB, MSSQL. Triggers: 'database', 'postgres', 'mysql', 'mongodb', 'schema', 'migration', 'pgbouncer', 'EXPLAIN'.
Use when adding, modifying, optimizing, or debugging CuTile autotuning code. Trigger signals: `exhaustive_search` / `replace_hints` / `hints_fn` / `cuda.tile.tune` in code, `autotune` in filenames, or correctness/performance issues in autotuned CuTile kernels. Covers: tune-once/cache/launch pattern, per-architecture configs (sm80–sm120), parameter space design (tile sizes, occupancy, num_ctas), and 7 common pitfalls with solutions.
Use when implementing Liquid Glass effects, reviewing UI for Liquid Glass adoption, debugging visual artifacts, optimizing performance, or requesting expert review of Liquid Glass implementation - provides comprehensive design principles, API patterns, and troubleshooting guidance from WWDC 2025. Includes design review pressure handling and professional push-back frameworks
PostgreSQL relational database. Covers SQL queries, indexes, constraints, and performance. Use when working with PostgreSQL. USE WHEN: user mentions "postgres", "postgresql", "pg_", asks about "JSONB queries", "window functions", "recursive CTE", "row level security", "full text search", "partitioning", "pgBouncer", "replication" DO NOT USE FOR: MySQL syntax - use `mysql` instead, MongoDB - use `mongodb` instead, Oracle PL/SQL - use `plsql` instead, SQL Server T-SQL - use `tsql` instead
NCU-driven iterative optimization workflow for CUDA/CUTLASS/Triton/CuTe DSL kernels. MANDATORY: every optimization MUST start with NCU profiling, followed by multi-dimensional analysis, then targeted code modification, then re-profiling to verify. Supports roofline, memory hierarchy, warp stalls, instruction mix, occupancy, divergence analysis. Provides implementation-specific code modifications: Native CUDA (launch config, memory patterns, async copy, Tensor Core), CUTLASS (ThreadblockShape, stages, epilogue, schedule policy, alignment), Triton (autotune params, compiler hints, tl.* API patterns), CuTe DSL (threads_per_cta, elems_per_thread, tiled_copy, copy atom, shared memory, warp/cta reduce). Use when optimizing any CUDA kernel performance.
Iteratively optimize cuTile kernel performance through systematic profiling, bottleneck analysis, IR comparison, and targeted tuning. Covers tile sizes, occupancy, autotune configs, TMA, latency hints, persistent scheduling, num_ctas, flush_to_zero, and IR-level debugging. Use when asked to "optimize cutile kernel", "improve kernel perf", "tune cutile performance", "make kernel faster", or iteratively benchmark and refine a cuTile GPU kernel in the TileGym project.
MoE expert-parallel communication overlap in Megatron Bridge. Covers dispatch/combine overlap, flex dispatcher backends, and expert wgrad scheduling.
Modern shell prompt configuration with Powerlevel10k and Zsh Vi Mode. Use when configuring shell prompts, setting up vi/vim keybindings in zsh, customizing cursor styles per mode, adding mode indicators, optimizing prompt performance, or troubleshooting slow prompts. Covers P10k instant prompt, vi mode plugins, and cursor customization.
Use when the user wants to push past conventional workflow limits with advanced performance techniques like parallel orchestration, streaming pipelines, or adaptive routing.
Generate Triton kernel code for Ascend NPU based on operator design documents. Used when users need to implement Triton operator kernels and convert requirement documents into executable code. Core capabilities: (1) Parse requirement documents to confirm computing logic (2) Design tiling partitioning strategy (3) Generate high-performance kernel code (4) Generate test code to verify correctness.
Universal Runtime best practices for PyTorch inference, Transformers models, and FastAPI serving. Covers device management, model loading, memory optimization, and performance tuning.
End-to-end SGLang SOTA performance workflow. Use when a user names an LLM model and wants SGLang to match or beat the best observed vLLM and TensorRT-LLM serving performance by searching each framework's best deployment command, benchmarking them fairly, profiling SGLang if it is slower, identifying kernel/overlap/fusion bottlenecks, patching SGLang code, and revalidating with real model runs.