MLOps
模型部署、评估与运维
ac-idea-lab
craigvandotcom
Use to deeply work a raw IDEA, framework, concept, or strategy (anything without execution steps yet) — two modes: GENIUS forensic first-principles critique (stress-test, find flaws, distill) and ALIEN paradigm transcendence (escape the frame, cross-domain transplants, expand). Triggers on "review this idea", "stress-test this", "devil's advocate", "critique this concept", "find flaws in this", "first-principles review", "transcend this idea", "go deeper", "push beyond analysis", "alien perspective", "escape local optima", "what am I missing at a deeper level". For a written implementation plan with steps/timelines use ac-plan-lab; for open-ended generation of new options use brainstorming; for a multi-model panel use multi-model.
jenkins-pipeline-generator
ehtbanton
Generate Jenkins pipeline files (Jenkinsfile) with declarative or scripted syntax. Triggers on "create jenkinsfile", "generate jenkins pipeline", "jenkins ci config", "jenkins build pipeline".
supabase-environments
MildTomato
Internal guide for building the Supabase CLI environments system. Covers three-environment model, variable resolution, pull/push workflows, secret handling, branch overrides, and local file conventions. For Supabase internal development - use when implementing the env CLI subsystem or environment variable infrastructure.
Meta-Learning Suite (MAML/FOMAML/Reptile + MAML++ Enhancements)
sovr610
This skill should be used when the user asks to "implement meta-learning", "add MAML inner loop", "implement FOMAML", "add Reptile", "implement MAML++", "add per-layer per-step learning rates", "implement LSLR", "add multi-step loss", "implement episodic sampling", "add few-shot learning", "implement inner-loop optimizer", "add second-order meta-gradients", "implement torch.func inner loop", "add higher library support", "implement adaptation curves", "add AUAC metrics", "implement meta-checkpoint format", "add derivative-order annealing", "implement episode sampler", "add Omniglot dataset", "add mini-ImageNet splits", "implement Phase 7 runner", "add meta-training loop", "implement fast adaptation", "add batch norm handling for meta-learning", "implement differentiable inner loop", or mentions MAML, FOMAML, Reptile, meta-gradients, episodic few-shot, inner-loop optimization, MAML++ enhancements, or Phase 7 meta-training in the cognitive pipeline.
filter-tune
alasdairnc
Run the CaseDive filter calibration pipeline in the correct sequence with baseline safety check. Checks for an existing baseline before calibrating, then compares results.
domain-layer
Nomik94
Domain Layer 설계 및 구현 가이드. Use when: Entity/Value Object/Aggregate Root 설계, 도메인 이벤트 구현, 비즈니스 로직 배치 판단, Repository Protocol 정의, Domain vs Application Service 구분, 상태 전이 로직, 도메인 예외 설계, 서비스 비대화 해결, 로직 분리. NOT for: 단순 CRUD (비즈니스 규칙 없으면 domain layer 불필요), SQLAlchemy 모델 작성, 단순 dataclass 문법.
cohere-java-sdk
RSHVR
Cohere Java SDK reference for chat, streaming, embeddings, reranking, and tool use. Use when building Java/Kotlin applications with Cohere APIs.
a0p-instancing
The-Interdependency
"Methodology for instancing agents in a0-betatest (the a0p research instrument), whose model diverges from canonical a0. Load this when adding or changing an AgentInstance / CharacterSheet CRUD path, a per-instance native ZFAE weight bank or its training/distillation loop, a ZFAE inference mode, a sentinel evaluation or pending-override gate, a per-agent safetensors checkpoint, or volatile sub-context memory — anywhere under a0-betatest backend/. Use it before writing code that creates, addresses, trains, runs, governs, or persists an a0p agent, so the code follows a0p's per-user CRUD + native-ZFAE + sentinel model instead of a0's spawn/fork/merge model. For canonical a0 and its mirror a0ucns, use agent-instantiation instead — a0p does NOT have sub_agent_spawn, a spawn executor, or InstanceMerge."
pricing-strategy
jamelna-apps
When the user mentions "pricing", "price", "monetization", "subscription", "tiers", "freemium", "revenue model", or asks about how to price a product or service.
SKILL: Документирование PuzzleAI
marozz1k2
документация docs.pxsto.re
deepagent
htooayelwinict
Expert guidance for DeepAgents framework - simplified agent creation with tool integration for LangChain/LangGraph workflows.
Gradient Checkpointing (Activation Recomputation)
sovr610
This skill should be used when the user asks to "enable gradient checkpointing", "reduce training memory", "activation checkpointing", "torch.utils.checkpoint", "memory-compute tradeoff", "checkpoint sequential layers", "selective checkpointing", "recomputation strategy", "activation memory profiling", "per-layer memory budget", "checkpoint_sequential", "checkpoint_wrapper", "SAC selective activation checkpointing", "SNN timestep checkpoint", "FSDP activation checkpointing", "checkpoint per timestep", "memory-efficient training", "recompute activations in backward", or needs guidance on trading compute for memory during training, per-layer memory profiling, selective recomputation strategies, or integration with distributed training wrappers.
cerebras-api
diskd-ai
Cerebras API integration for building AI-powered applications with ultra-fast LLM inference. Use when working with Cerebras's Chat Completions API, Python SDK (cerebras_cloud_sdk), TypeScript SDK (@cerebras/cerebras_cloud_sdk), tool use/function calling, structured outputs with JSON schemas, reasoning models with thinking tokens, streaming responses, or any Cerebras API integration task. Triggers on mentions of Cerebras, Cerebras Inference, Llama on Cerebras, Qwen on Cerebras, GLM, or fast LLM inference needs.
hreng-skills
clous-ai
Use when the user wants to assess engineering team skills, build a skills matrix, identify gaps vs. roadmap, and design training or hiring plans.
plan
pmco23
Use after /review to transform the approved design into an atomic execution plan. Writes task groups with exact file paths, complete code examples, and named test cases with assertions. Build agents must never need to ask clarifying questions. Writes .pipeline/plan.md.
eval-frameworks
cuba6112
Evaluation framework patterns for RAG and LLMs, including faithfulness metrics, synthetic dataset generation, and LLM-as-a-judge patterns. Triggers: ragas, deepeval, llm-eval, faithfulness, hallucination-check, synthetic-data.
pipeline-pattern-react
progmichaelkibenko
Implements the Pipeline design pattern in React for data transformation. Use when the user mentions pipeline pattern, or when you need a fixed sequence of stages that each transform data and pass to the next—ETL-style processing in the UI, parsing, formatting pipelines, or any linear transformation flow that runs to completion.
ecostral-optimizer
wilfred-dore
Optimize any HuggingFace model to minimize energy consumption, CO₂ emissions, and inference cost using real GPU measurements and Mistral Large reasoning. Use this skill when asked to quantize a model, reduce its carbon footprint, benchmark deployment configurations (datacenter GPU, Jetson Orin, edge devices), or generate an optimization report with W&B experiment tracking.
cmev-prepare-dataset
RodelFMiguel
Inspect and prepare CLAIM-CMEV image, document, or repair-cost data for reproducible experiments, including provenance, label mapping and grouped splits. Use for dataset onboarding or conversion, not model fitting.
agent-instantiation
The-Interdependency
Methodology for instantiating, forking, running, merging, and retiring agents in the a0 platform and its near-identical mirror a0ucns. Load this when adding or changing a sub-agent spawn path, a PCNA instance fork/merge, an agent definition or naming scheme, spawn caps or approval gating, an agent run/log table, a heartbeat-driven agent task, or a checkpoint of agent state. Use it before writing any code that creates, addresses, schedules, or tears down an agent or sub-agent, so the new code follows the platform's existing lifecycle, fork/merge, identity, and gating contracts rather than inventing a parallel one. NOTE: a0-betatest (a0p) has diverged to a different per-user CRUD + native-ZFAE instancing model — this skill's spawn/fork/merge sequence does NOT apply there; see "a0-betatest divergence".
ac-plan
craigvandotcom
'Turn an idea into ONE ac2 plan file — problem, approach, deliverables, assumptions, risk + sequence, out-of-scope, and a success criterion that can come out FALSE. Explorers optional, chosen by size. Triggers: "ac2 plan", "write an ac2 plan", "plan this for ac2". Hands off to ac-polish.'
django-model
jprokay-counterpart
Creating Django models following Counterpart's patterns. Use when building new models, working with audit fields, implementing relationships, or using PydanticJSONEncoder. Triggers: 'create model', 'new database table', 'audit fields', 'BaseModel inheritance', 'add foreign key', 'JSON field validation'
Data Loader Throughput + Sequence Packing
sovr610
This skill should be used when the user asks to "audit dataloader throughput", "measure data pipeline stalls", "implement sequence packing", "reduce padding waste in training", "add cu_seqlens packing for SFT", "boundary-aware packing FlashAttention", "DataCollatorWithFlattening", "configure DataLoader num_workers prefetch_factor", "tune DataLoader settings", "streaming dataset pipeline", "WebDataset tar shards", "memmap pretraining dataset", "HF datasets streaming IterableDataset", "deterministic sharding per rank", "DistributedSampler set_epoch", "padding ratio metrics", "effective tokens per second", "pretraining block builder", "data stall ratio measurement", "persistent_workers pin_memory tuning", "shard cache policy", "packed sequences with position_ids reset", "varlen_attn cu_seqlens", or needs guidance on turning the input pipeline into a first-class performance target with measurement, streaming I/O, packing, and deterministic distributed sharding.
bg-remover
yemyat
Remove backgrounds from images using FAL.ai's BiRefNet model. Use when users ask to remove background, make transparent PNG, extract subject from image, or create cutouts. Trigger phrases include "remove background", "transparent background", "cut out", "extract subject", or any background removal request.