Data & AI skills

Onboarding-and-switching playbook for SMB Product-Builder products. Defines the first-run experience that turns a prospect leaving an incumbent (ServiceTitan/Toast/Mindbody/Shopify/QuickBooks) into an activated user — import-first onboarding, the activation milestone, sample-data fallback, and the time-to-first-value target. Applied by migration-import-engineer and architect/pm so onboarding is designed as a funnel, not an afterthought. Our whole wedge is "low switching cost"; this skill makes that real on day one.

The English summary is being prepared. Showing the original description.

Store data permanently in the Cognee knowledge graph. Accepts a data category (user, project, or agent) to tag the data with the correct node_set for filtered retrieval.

The English summary is being prepared. Showing the original description.

Search Cognee memory. Session memory is automatically searched on every prompt via hooks. Use this skill explicitly for permanent knowledge graph search, filtered category search, or when you need more results than the automatic lookup provides.

The English summary is being prepared. Showing the original description.

Forget data stored in Cognee memory. Use when the user asks to forget, delete, or remove something from memory (e.g. "forget what we talked about tennis", "delete that from memory"). Finds the documents in the current dataset whose raw content matches what the user wants forgotten and deletes them.

The English summary is being prepared. Showing the original description.

Applies production AI integration patterns for chat, structured output, guardrails, provider routing, and AI UX. Use when adding LLM-powered features to an application.

The English summary is being prepared. Showing the original description.

Designs product and site search for end users. Use when choosing search engines, indexing, reindexing, relevance tuning, facets, autocomplete, merchandising, or search analytics.

The English summary is being prepared. Showing the original description.

Tests whether data supports a claim: A/B significance, peeking, sample size, eval error bars, multiple comparisons. Use when deciding if results are significant.

The English summary is being prepared. Showing the original description.

Builds local-first note vault retrieval for Obsidian, markdown notebooks, and NotebookLM-adjacent exports. Use when packaging notes into LLM-ready context.

The English summary is being prepared. Showing the original description.

Proves protocols and agent workflows with TLA+, P, Alloy, Z3, Dafny, or Lean. Use when proving protocols correct, writing TLA+ specs, or verifying LLM proofs.

The English summary is being prepared. Showing the original description.

Evaluates neuromarketing studies and neural-data law. Use when judging EEG, fMRI, eye-tracking, or GSR ad study or vendor, or if signals are regulated neural data.

The English summary is being prepared. Showing the original description.

Automates Metabase cards, dashboards, Remote Sync, embedding, tenants, and the Agent API/MCP server for AI workflows. Use when scripting, promoting, or embedding Metabase content.

The English summary is being prepared. Showing the original description.

Chooses among AI/ML approaches: classical ML, LLM, RAG, fine-tuning, agents, embeddings, MoE or SSM. Use when picking or scaling an architecture, or self-host vs API.

The English summary is being prepared. Showing the original description.

Builds LLM pretraining corpora from Common Crawl: extract, filter, deduplicate, decontaminate, mix, synthesize. Use when curating or ablating a pretraining corpus or data pipeline.

The English summary is being prepared. Showing the original description.

Builds ML pipelines: EDA, features, leakage checks, evaluation. Use when doing data science or explaining fairness, privacy, speech, vision-language, or diffusion mechanics.

The English summary is being prepared. Showing the original description.

Builds and backtests demand and revenue forecasts: rolling-origin backtests, prediction intervals, reconciliation, zero-shot foundation models. Use when forecasting time series.

The English summary is being prepared. Showing the original description.

Designs operating models for AI in product teams. Use when planning platform ownership, provider strategy, data boundaries, evals, or sensitive-data controls.

The English summary is being prepared. Showing the original description.

Prompt engineering for production LLMs — structured outputs, evals, RAG, tool workflows, multimodal prompting, and safety. Use when designing, debugging, or shipping prompts.

The English summary is being prepared. Showing the original description.

Builds analytics engineering layers for metrics, contracts, and BI-ready models. Use when shaping dbt or SQLMesh marts, metric governance, lineage, or data quality.

The English summary is being prepared. Showing the original description.

Designs streaming platforms for Kafka, Flink, CDC, and lakehouse ingestion. Use when planning event backbones, CDC pipelines, schema governance, or real-time lakehouse delivery.

The English summary is being prepared. Showing the original description.

Builds multi-repo context hubs and compiled markdown knowledge maps. Use when profiling repo portfolios or assembling LLM-ready cross-repo knowledge bases.

The English summary is being prepared. Showing the original description.

Designs lakehouse platforms across Iceberg, Delta, Hudi, and Paimon. Use when choosing catalogs, CDC paths, query engines, governance, or cost controls.

The English summary is being prepared. Showing the original description.

Adapts LLMs: SFT dataset loss masking, LoRA and QLoRA, full FT vs PEFT, distillation, pruning, tokenizer fragmentation. Use when fine-tuning or compressing a model.

The English summary is being prepared. Showing the original description.

LLM serving for latency, batching, caching, quantization, and routing. Use when tuning vLLM or SGLang, sizing KV cache, adding speculative decoding, or cutting serving cost.

The English summary is being prepared. Showing the original description.

Designs RAG retrieval for LLM grounding: chunking, BM25+vector hybrid, reranking, GraphRAG. Use when adding embeddings to SQLite FTS5 search or running held-out retrieval eval.

The English summary is being prepared. Showing the original description.