Agent Skills

---

name: nemo-guardrails

description: NVIDIA's runtime safety framework for LLM applications. Features jailbreak detection, input/output validation, fact-checking, hallucination detection, PII filtering, toxicity detection. Uses Colang 2.0 DSL for programmable rails. Production-ready, runs on T4 GPU.

---

name: nnsight-remote-interpretability

description: Provides guidance for interpreting and manipulating neural network internals using nnsight with optional NDIF remote execution. Use when needing to run interpretability experiments on massive models (70B+) without local GPU resources, or when working with any PyTorch architecture.

---

name: openrlhf-training

description: 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× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.

---

name: peft-fine-tuning

description: Parameter-efficient fine-tuning for LLMs using LoRA, QLoRA, and 25+ methods. Use when fine-tuning large models (7B-70B) with limited GPU memory, when you need to train <1% of parameters with minimal accuracy loss, or for multi-adapter serving. HuggingFace's official library integrated with transformers ecosystem.

---

name: prompt-guard

description: Meta's 86M prompt injection and jailbreak detector. Filters malicious prompts and third-party data for LLM apps. 99%+ TPR, <1% FPR. Fast (<2ms GPU). Multilingual (8 languages). Deploy with HuggingFace or batch processing for RAG security.

---

name: pyvene-interventions

description: Provides guidance for performing causal interventions on PyTorch models using pyvene's declarative intervention framework. Use when conducting causal tracing, activation patching, interchange intervention training, or testing causal hypotheses about model behavior.

---

name: ray-data

description: Scalable data processing for ML workloads. Streaming execution across CPU/GPU, supports Parquet/CSV/JSON/images. Integrates with Ray Train, PyTorch, TensorFlow. Scales from single machine to 100s of nodes. Use for batch inference, data preprocessing, multi-modal data loading, or distributed ETL pipelines.

---

name: rwkv-architecture

description: RNN+Transformer hybrid with O(n) inference. Linear time, infinite context, no KV cache. Train like GPT (parallel), infer like RNN (sequential). Linux Foundation AI project. Production at Windows, Office, NeMo. RWKV-7 (March 2025). Models up to 14B parameters.

---

name: sentencepiece

description: Language-independent tokenizer treating text as raw Unicode. Supports BPE and Unigram algorithms. Fast (50k sentences/sec), lightweight (6MB memory), deterministic vocabulary. Used by T5, ALBERT, XLNet, mBART. Train on raw text without pre-tokenization. Use when you need multilingual support, CJK languages, or reproducible tokenization.

---

name: simpo-training

description: Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model needed, more efficient than DPO. Use for preference alignment when want simpler, faster training than DPO/PPO.

---

name: slime-rl-training

description: Provides guidance for LLM post-training with RL using slime, a Megatron+SGLang framework. Use when training GLM models, implementing custom data generation workflows, or needing tight Megatron-LM integration for RL scaling.

---

name: sparse-autoencoder-training

description: Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable features. Use when discovering interpretable features, analyzing superposition, or studying monosemantic representations in language models.

---

name: torchforge-rl-training

description: Provides guidance for PyTorch-native agentic RL using torchforge, Meta's library separating infra from algorithms. Use when you want clean RL abstractions, easy algorithm experimentation, or scalable training with Monarch and TorchTitan.

---

name: transformer-lens-interpretability

description: Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing activation patching experiments.

---

name: unsloth

description: Expert guidance for fast fine-tuning with Unsloth - 2-5x faster training, 50-80% less memory, LoRA/QLoRA optimization

---

name: verl-rl-training

description: Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.

---

name: add-memory-kind

description: Add a new business memory kind end-to-end. Pick the storage combination (Markdown / SQLite / LanceDB), pick the markdown strategy (daily-log / skill-named / single-file), then wire up the schema(s), repo(s), and writer(s).

---

name: commit

description: Stage and create a Conventional Commits message for the current change

---

name: new-branch

description: Create a GitHub branch from main with the project naming convention

---

name: pr

description: Open a GitHub PR targeting the correct branch with the project template

---

name: release

description: Cut a versioned release and publish everos to PyPI via the tag-triggered workflow

---

name: using-ao

description: Catalog of the AO (Agent Orchestrator) `ao` CLI: spawning workers, managing sessions and projects, creating reusable Cues, sending messages, controlling the shared browser, previewing pages, and daemon control. Use when using the ao CLI, creating Cues, spawning workers, or managing AO sessions in an AO workspace.

---

name: using-ao

description: Catalog of the AO Cloud `ao` CLI available inside a cloud sandbox: spawning worker sessions, listing them, messaging them, reporting to the orchestrator, killing them, claiming pull requests, and starting AO reviews. Use when orchestrating or coordinating AO cloud sessions with the ao CLI.

---

name: beautiful-article

description: 把用户提供的素材（网页 URL / PDF / DOCX / Markdown / 纯文本 / 截图 / 粘贴材料）编辑、设计成一篇美丽的、可离线打开和分享的**单文件 HTML 网页文章**。基于 reacticle 组件协议：不手写裸 HTML/CSS，而用语义组件 + 受主题约束的 Raw 自由层；按 source→规划→双确认→生成→终审→修复的小型 harness 流程推进，默认 100% 信息保留的长文。触发场景：把 URL/PDF/DOCX/文章做成网页文章 / 长文 / briefing / 解释文 / 视觉文章 / 教程 / 审阅复盘 / 方案分析，'render this as a beautiful web article / 把这篇做成网页文章 / 生成一篇可分享的 HTML 长文 / reacticle 文章'。只生成文章，不生成后台、表单、dashboard、产品原型或通用 Web App。