MLOps
模型部署、评估与运维
dotfile-audit
harperreed
Use when you have an existing pipeline DOT file that needs review, fixing, or validation. Use when validation produces warnings or errors, when a DOT file was hand-written or generated by another tool, or when you want to verify a pipeline covers 100% of a spec before running it.
peft-fine-tuning
majiayu000
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.
veo3-1
delneg
Generate videos with Google Veo 3.1 via grsai.com API. Use for any video creation, text-to-video, or motion content requests. Triggers on "generate video", "create a clip", "make a video of", "video of", "animate". Supports fast/pro quality tiers; 720p/1080p/4K; 4/6/8-second durations; 16:9 and 9:16 aspect ratios.
start
TakaGoto
"Begin your RAG learning journey — assess your level and pick a learning path"
naval-almanack
kangarooking
当用户询问《纳瓦尔宝典》这本书本身(书名/作者/章节/整书概览/术语)，或涉及本书未独立成 Skill 的主题时调用: 幸福训练/欲望管理、接受与内耗、冥想与心猴、地位游戏与攀比、选朋友与圈子、打工vs创业(委托代理)、 诚实沟通、身份与自我形象、期望与愤怒、生命意义、该不该信玄学(理性验证)、长期复利与声誉、判断力训练。 以下意图优先使用同包独立 Skill: wealth-structure(致富/股权期权)、productize-yourself(职业定位/副业方向)、 decision-heuristics(重大选择纠结)、hourly-rate-time(时薪/外包/时间分配)、reading-metaskill(怎么读书学习)、screen-detox(刷手机上瘾)。 Triggers: 纳瓦尔/Naval/纳瓦尔宝典/幸福/冥想/内耗/攀比/圈子/复利/人生意义
rational-buddhism
kangarooking
当用户问某主张/玄学/灵修该不该信、想建立自己的验证标准、或寻找科学与精神的结合点时调用。 核心理念: 尝试一切, 亲自测试, 保持怀疑, 保留有用的, 舍弃没用的; 不可证伪的主张不进入基本真理; 冥想/接纳等内在技术可保留。 不适用于: 需要信仰安慰的临终/哀伤场景(尊重个体选择)。 Triggers: 该不该信/玄学/灵修/冥想/可证伪/验证/理性/rational buddhism/verify/belief
cohere-go-sdk
RSHVR
Cohere Go SDK reference for chat, streaming, embeddings, reranking, and tool use. Use when building Go applications with Cohere APIs.
brief
pmco23
Use when starting a new feature, project, or task - extracts requirements, constraints, non-goals, style preferences, and key concepts from fuzzy input through conversational Q&A. Output is a structured brief saved to .pipeline/brief.md. Always run this before /design.
slime-rl-training
majiayu000
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.
simpo-training
majiayu000
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.
audio-transcribe
maxgent-ai
Speech-to-text transcription using Whisper with word-level timestamps. Use when users ask to transcribe audio or video to text, generate subtitles, or recognize speech.
hughes-property-based-testing
copyleftdev
Test software in the style of John Hughes, inventor of QuickCheck and property-based testing. Emphasizes specifying properties that should hold for all inputs, generating random test cases, and shrinking failures to minimal examples. Use when testing algorithms, data structures, parsers, serializers, or any code with clear invariants.
oh-my-pi
Changhochien
"Multi-agent orchestration for PI: team coordination, autonomous execution, persistence loops, and requirements gathering. USE FOR: multi-agent development, parallel execution, code implementation, architecture design, security review, test automation, project planning. Commands: /skill:team, /skill:autopilot, /skill:ralph, /skill:deep-interview, /skill:plan, /skill:verify"
ai-engineering-skill
odewahn
Practical guide for building production ML systems based on Chip Huyen's AI Engineering book. Use when users ask about model evaluation, deployment strategies, monitoring, data pipelines, feature engineering, cost optimization, or MLOps. Covers metrics, A/B testing, serving patterns, drift detection, and production best practices.
fabric-icons
AlahmadiQ8
Search and fetch official Microsoft Fabric / Azure SVG icons for use in diagrams, documentation, and UI. Use when building architecture diagrams (HTML, PPTX, or any visual format) that need icons for Fabric concepts like Lakehouse, Data Warehouse, Pipeline, Eventhouse, Power BI, OneLake, Dataflow, Notebook, Semantic Model, KQL Database, and other Fabric items. Also use when the user asks for Microsoft Fabric or Azure data platform icons.
nvidia-specdecode-al-doctor
IgorGanapolsky
Fail-closed speculative-decoding AL/D doctor for ThumbGate. Checks speedup ≤ AL/(1+ρD), attention D=128/G-1, tile alignment, and maps to checkpoint-speculative-decoding-acceptance. Steal from NVIDIA co-design blog — never clone TensorRT/EAGLE/Model-Optimizer. Slash: /nvidia-specdecode-al-doctor.
experiment-tracking-swanlab
majiayu000
Provides guidance for experiment tracking with SwanLab. Use when you need open-source run tracking, local or self-hosted dashboards, and lightweight media logging for ML workflows.
blecsd-3d
Kadajett
Render 3D graphics in the terminal with @blecsd/3d. Covers software rasterizer, 3D math (vec3/mat4), multiple backends (braille/sixel/kitty), OBJ loading, scene graphs, and camera systems.
z-cheatsheet-add
AutoFor
z-cheatsheet にエントリを追加する。
senior-computer-vision
nimeshgurung
World-class computer vision skill for image/video processing, object detection, segmentation, and visual AI systems. Expertise in PyTorch, OpenCV, YOLO, SAM, diffusion models, and vision transformers. Includes 3D vision, video analysis, real-time processing, and production deployment. Use when building vision AI systems, implementing object detection, training custom vision models, or optimizing inference pipelines.
security-review
pmco23
Use after build is complete to scan for OWASP Top 10 vulnerabilities. Checks injection, authentication, authorization, data exposure, and misconfiguration risks. Requires .pipeline/build.complete.