Browse Skills

Provides guidance for LLM post-training with RL using slime, a Megatron+SGLang framework. Use when training GLM models, implementing custom data generation work

Specialized skill for building production-ready serverless applications on GCP. Covers Cloud Run services (containerized), Cloud Run Functions (event-driven), c

draw.io diagram creation, editing, and review. Use for .drawio XML editing, PNG conversion, layout adjustment, and AWS icon usage.

High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× fast

Query, stage, and apply configuration changes for Railway environments. Use for ANY variable or env var operations, service configuration (source, build setting

Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throu

Generate descriptive commit messages by analyzing git diffs. Use when the user asks for help writing commit messages or reviewing staged changes.

Automate Docker Hub tasks via Rube MCP (Composio): repositories, images, tags, and container registry management. Always search tools first for current schemas.

Server management principles and decision-making. Process management, monitoring strategy, and scaling decisions. Teaches thinking, not commands.

Optimizes LLM inference with NVIDIA TensorRT for maximum throughput and lowest latency. Use for production deployment on NVIDIA GPUs (A100/H100), when you need

Production deployment principles and decision-making. Safe deployment workflows, rollback strategies, and verification. Teaches thinking, not scripts.

Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms fo