MLOps
模型部署、评估与运维
kgz — Execute Code on Kaggle Kernels (GPU & TPU)
mlnomadpy
Dep: websocket-client
huggingface-daemon
fast-gateway-protocol
Fast Hugging Face Inference API via FGP daemon. Use when user needs text generation, embeddings, classification, image captioning, or model inference. Triggers on "huggingface inference", "generate text", "get embeddings", "classify text", "hugging face", "HF model", "zero-shot".
embedding-comparison
mindmorass
Compare and evaluate embedding models for semantic search
finance-forecast
jforksy
Detailed financial scenario modeling, revenue projections, and burn rate analysis
explaining
AlexanderStephenThompson
Repeatable execution process for producing clear explanations. Covers Subject and Situational frameworks, depth scaling, and relatability tools.
evaluation-metrics
pluginagentmarketplace
LLM evaluation frameworks, benchmarks, and quality metrics for production systems.
Four Forces of Progress
Coowoolf
A behavioral model defining the opposing forces in switching decisions—Push, Pull, Anxiety, Habit. Change happens only when (Push + Pull) > (Anxiety + Habit). Core to Jobs-to-be-Done theory.
ai-vision-go
httprunner
Multimodal UI understanding and single-step planning via OpenAI-compatible Responses APIs. Use when you need AIQuery/AIAssert and plan-next to extract UI element coordinates, validate UI assertions, summarize screenshots, or decide the next UI action from an image. External agents handle execution via adb/hdc and multi-step loops. Defaults to Doubao models but can be pointed at other multimodal providers via base URL, API key, and model name.
model-usage
mmcmedia
Use CodexBar CLI local cost usage to summarize per-model usage for Codex or Claude, including the current (most recent) model or a full model breakdown. Trigger when asked for model-level usage/cost data from codexbar, or when you need a scriptable per-model summary from codexbar cost JSON.
agentscope
changxubo
Build production-ready multi-agent systems with AgentScope framework. Use this skill when users want to create AI agents, multi-agent workflows, or distributed agent applications. Provides patterns for ReActAgent, MsgHub, Pipelines, MCP integration, memory management, and fault-tolerant agent systems.
J-Curve Career Framework
Coowoolf
High-growth careers are J-Curves, not stairs—you jump off a cliff (take risk), struggle for 6-9 months (bottom of J), then shoot up exponentially. Use when deciding between safe promotion vs stretch role.
Corporate Innovation C-Corp Model
Coowoolf
To replicate startup speed in large companies, launch new products as separate legal entities (C-Corps) with distinct brands, reporting directly to the CEO, bypassing standard chains of command.
llm-basics
pluginagentmarketplace
LLM architecture, tokenization, transformers, and inference optimization. Use for understanding and working with language models.
renaissance-statistical-arbitrage
copyleftdev
Build trading systems in the style of Renaissance Technologies, the most successful quantitative hedge fund in history. Emphasizes statistical arbitrage, signal processing, and rigorous scientific methodology. Use when developing alpha research, signal extraction, or systematic trading strategies.
maciver-hypothesis-testing
copyleftdev
Test software in the style of David MacIver, creator of Hypothesis. Emphasizes sophisticated shrinking, example databases, stateful testing, and practical property-based testing in Python. Use when testing Python code with complex data structures, APIs, or stateful systems.
modelscope-notebook-adapter
VoyagerXvoyagerx
Adapt and reproduce Jupyter notebooks, Colab notebooks, GitHub tutorial notebooks, or local notebook folders in ModelScope Notebook. Use when Codex must port a tutorial to ModelScope, replace Hugging Face/Google Drive/Colab downloads with ModelScope model or dataset repositories, prepare/upload missing assets with README metadata and license/source information, execute notebooks locally or in ModelScope-compatible environments, debug dependency/data/model issues, and update reports with reproducibility results.
lets_build_an_llm Skill
7Hu0v0
How does this technical point help me judge a candidate's real capability, team ownership, and talent value before a pre-talk?
fowler
copyleftdev
Design systems using Martin Fowler's principles of refactoring, continuous integration, and patterns of enterprise application architecture. Emphasizes clean code, evolution over revolution, and writing code for humans first. Use when designing enterprise systems, planning refactors, or establishing engineering culture.
peer-review
pietz
Ask a group of LLM tools for their opinion on a given task. Use this when stuck on a problem, seeking a second opinion, validating an approach, or wanting diverse viewpoints on complex decisions.
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.
python-immutable-accumulator
shimo4228
"Use when building immutable state accumulators in Python. Frozen dataclass + tuple pattern with slots gotcha."
mixseek-config-validate
drillan
MixSeekの設定ファイル（team.toml、orchestrator.toml、evaluator.toml、judgment.toml）を検証します。「設定を検証」「TOMLをチェック」「設定ファイルの確認」「バリデーション」「ワークスペースの検証」といった依頼で使用してください。TOML構文とMixSeekスキーマへの準拠を確認します。
alliance-ml
ayushnangia
"Guide for running ML/AI workloads on Alliance Canada (formerly Compute Canada) HPC clusters. Use this skill whenever the user mentions Alliance Canada, Compute Canada, Narval, Cedar, Graham, Trillium, Niagara, Fir, Nibi, Rorqual, Killarney, Slurm job submission on Canadian clusters, sbatch for ML training, GPU jobs on HPC, virtualenv on clusters, or anything related to running deep learning on shared computing infrastructure. Also use when the user asks about CUDA on clusters, distributed training with DeepSpeed/PyTorch DDP on Slurm, managing datasets on parallel filesystems, HuggingFace on HPC, Apache Arrow or Parquet on clusters, ML best practices for shared computing, Apptainer, Singularity, containers on HPC, vLLM, inference serving on HPC, or LLM inference on clusters."
state-modeling
velcrafting
Define client state shape and transitions for non-trivial UI flows (multi-step, branching, async, derived state).