Java language (17+). Covers modern features, patterns, and best practices. Use when writing Java applications, Spring Boot backends, or enterprise systems. USE WHEN: user mentions "java", "records", "sealed classes", "streams", asks about "pattern matching", "switch expressions", "Optional", "collections", "generics" DO NOT USE FOR: Spring Boot specifics - use `backend-spring-boot` skill instead DO NOT USE FOR: Lombok annotations - use `lombok` skill instead DO NOT USE FOR: MapStruct - use `mapstruct` skill instead

LangGraph state machines for RAG and agentic flows. Covers typed state, conditional edges for routing (answer/clarify/retrieve/rewrite), checkpointing with SqliteSaver/PostgresSaver, human-in-the-loop interrupts, multi-agent supervisor patterns, Self-RAG and CRAG as explicit graphs, combining with LangChain retrievers. USE WHEN: user mentions "LangGraph", "StateGraph", "agentic RAG", "conditional edges", "checkpointer", "human in the loop", "Self-RAG", "CRAG", "corrective RAG", "supervisor agent", "LangGraph RAG" DO NOT USE FOR: plain LangChain chains - use `langchain`; LlamaIndex workflows - use `llamaindex`; generic agentic RAG theory - use `agentic-rag`; DSPy programs - use `dspy`

LLM gateways in front of RAG stacks. Covers Portkey (caching, fallbacks, retries, observability), OpenRouter (300+ model routing), LiteLLM Proxy, Kong AI Gateway, semantic caching at gateway layer, cost-based routing (cheap model for easy queries), rate-limit handling, and unified API across providers. Config examples. USE WHEN: user mentions "LLM gateway", "Portkey", "OpenRouter", "LiteLLM", "Kong AI Gateway", "AI gateway", "semantic cache gateway", "provider fallback", "unified LLM API" DO NOT USE FOR: in-app caching - use `rag-caching`; multi-region routing - use `multi-region`; cost tracking/dashboarding - use `cost-allocation`

Multi-region RAG deployments for latency and resilience. Covers geo-replicated vector stores (Pinecone multi-region, Qdrant cluster, MongoDB Atlas Global, Weaviate), per-region embedding/rerank pools, LLM routing to nearest provider region (Anthropic, OpenAI, Bedrock), eventual-consistency strategies for index updates, and failover patterns. USE WHEN: user mentions "multi-region RAG", "global RAG", "geo replication", "RAG failover", "latency routing", "regional LLM endpoint", "disaster recovery RAG" DO NOT USE FOR: single-region scaling - use `rag-production`; LLM gateway vendor routing - use `llm-gateway`; CDN caching - use `rag-caching`

MySQL relational database. Covers queries, indexes, and optimization. Use when working with MySQL databases. USE WHEN: user mentions "mysql", "mariadb", asks about "AUTO_INCREMENT", "ON DUPLICATE KEY UPDATE", "GROUP_CONCAT", "mysql specific syntax" DO NOT USE FOR: PostgreSQL - use `postgresql` instead, MongoDB - use `mongodb` instead, Oracle - use `oracle` instead, SQL Server - use `sqlserver` instead

NATS cloud-native messaging system. Covers Core NATS, JetStream persistence, and request/reply patterns. Use for lightweight, high-performance microservices communication. USE WHEN: user mentions "nats", "jetstream", "cloud-native messaging", "request/reply", "subject wildcards", asks about "lightweight messaging", "microservices communication", "nats streaming" DO NOT USE FOR: complex routing - use `rabbitmq`; AWS-native - use `sqs`; Azure-native - use `azure-service-bus`; JMS compliance - use `activemq`; persistent queues only - use dedicated broker

Ontology-aware RAG. Leveraging domain ontologies (SNOMED CT, FIBO, schema.org, Gene Ontology, custom OWL) to expand queries with synonyms, hyponyms, and broader concepts; boost retrieval for ontology-matched entities; hybrid SPARQL + vector pipelines; RDFLib + embeddings. Examples with medical and financial ontologies. USE WHEN: user mentions "ontology", "SNOMED", "FIBO", "schema.org", "OWL", "SKOS", "RDF", "SPARQL RAG", "taxonomy-guided", "concept expansion", "hyponym retrieval" DO NOT USE FOR: building a KG from text - use `knowledge-graph-construction`; deduping entities - use `entity-resolution`; generic graph RAG - use `graph-rag`

Pinecone advanced features. Serverless vs pods architecture, hybrid search with sparse-dense vectors, namespaces for multi-tenancy, metadata filtering performance, upsert batching limits, collections (snapshot/restore), Pinecone-integrated inference, Python SDK v5+, Assistants API. USE WHEN: user mentions "Pinecone", "Pinecone serverless", "Pinecone namespaces", "sparse-dense Pinecone", "Pinecone Assistants", "Pinecone inference" DO NOT USE FOR: other managed vector DBs - use `vector-stores/milvus`, `vector-stores/mongodb-atlas-vector`, `vector-stores/pinecone-advanced` siblings; hybrid fusion math - use `rag/hybrid-search`

Python performance profiling with cProfile, tracemalloc, and line_profiler. Use for identifying bottlenecks and memory issues. USE WHEN: user mentions "Python profiling", "cProfile", "memory profiling", asks about "Python performance", "tracemalloc", "line_profiler", "py-spy", "Python optimization", "Python memory leak" DO NOT USE FOR: Java/Node.js profiling - use respective skills instead

Python security patterns for Django, FastAPI, and Flask. Covers Bandit, Safety, secure coding practices, and OWASP for Python ecosystem. USE WHEN: user works with "Python", "Django", "FastAPI", "Flask", asks about "Python vulnerabilities", "pip security", "Bandit", "Python injection", "Python authentication" DO NOT USE FOR: general OWASP concepts - use `owasp` or `owasp-top-10` instead, Node.js/Java security - use language-specific skills

LLM-as-reranker patterns. RankGPT listwise / pairwise / pointwise prompting, sliding window for long candidate lists, cost-latency vs Cohere Rerank, calibration tricks, structured output for rank lists. Code uses the Anthropic SDK. USE WHEN: user mentions "LLM reranker", "RankGPT", "listwise reranking", "pairwise reranking", "pointwise reranking", "GPT reranker", "Claude reranker" DO NOT USE FOR: standard cross-encoder reranking - use `rag/reranking`; fine-tuning cross-encoders - use `retrieval/cross-encoder-training`; ColBERT - use `retrieval/colbert-retrieval`

Spring Cloud Function for serverless and FaaS deployments. Covers function beans, AWS Lambda, Azure Functions, and GCP adapters. USE WHEN: user mentions "spring cloud function", "serverless Spring", "AWS Lambda Spring", "Azure Functions Spring", "FaaS", "function composition" DO NOT USE FOR: simple REST APIs - use standard controllers, complex workflows - consider Step Functions or similar

Declarative HTTP client for microservices communication with Spring Cloud OpenFeign. Covers @FeignClient, error handling, interceptors, and circuit breaker integration. USE WHEN: user mentions "feign", "openfeign", "@FeignClient", "declarative HTTP client", "service-to-service communication", "microservices client" DO NOT USE FOR: external APIs - consider WebClient or RestClient, simple HTTP calls - use RestClient

Spring for GraphQL - building GraphQL APIs with Spring Boot. Covers queries, mutations, subscriptions, @BatchMapping, DataLoader, and security. USE WHEN: user mentions "spring graphql", "@QueryMapping", "@MutationMapping", "@SubscriptionMapping", "@BatchMapping", "GraphQL Spring Boot", "N+1 GraphQL" DO NOT USE FOR: REST APIs - use standard Spring MVC, standalone GraphQL - use `graphql-java` skill

Spring for Apache Kafka integration with KafkaTemplate and @KafkaListener. Covers producers, consumers, retry topics, DLT, and transactions. USE WHEN: user mentions "spring kafka", "KafkaTemplate", "@KafkaListener", "kafka producer Spring", "kafka consumer Spring", "retry topic" DO NOT USE FOR: raw Kafka - use `kafka` skill, RabbitMQ - use `spring-amqp` instead

High-performance serving of embedding and reranker models in production. Covers HuggingFace Text Embeddings Inference (TEI) on GPU, Docker deployment, ONNX/FP16 quantization, dynamic batching, NVIDIA Triton for co-serving embedding + reranker + LLM, and NVIDIA NIM as a managed alternative. Benchmarks and configs. USE WHEN: user mentions "TEI", "Text Embeddings Inference", "Triton", "Triton Inference Server", "NIM", "NVIDIA NIM", "embedding server", "reranker server", "ONNX embedding", "serve BGE", "serve E5" DO NOT USE FOR: vector DB serving - use `rag-architecture`; LLM serving in general - use provider-specific docs; cost tracking - use `cost-allocation`

Temporal awareness in RAG. Recency bias weighting, exponential decay scoring, time-range filtering, LLM-based date extraction ("last quarter", "yesterday"), temporal knowledge graphs, freshness vs authority tradeoff, event-based retrieval, timestamp metadata, differential reindexing of old vs fresh docs. USE WHEN: user mentions "time-aware RAG", "recency bias", "temporal retrieval", "freshness", "date filter", "temporal knowledge graph", "exponential decay" DO NOT USE FOR: static metadata filters generally - use `self-querying-retriever`; CDC for ingestion - use `cdc-streaming-ingestion`; evaluation of time-sensitive answers - use `rag-evaluation`

Vespa hybrid retrieval and ranking. BM25 + ANN + ColBERT late interaction in one query, ranking phases (match-phase, first-phase, second-phase, global-phase), tensor expressions, cloud vs self-hosted, schema definition (.sd files), embedding in-cluster, scale to billions. USE WHEN: user mentions "Vespa", "Vespa schema", "Vespa ranking", "vespa.ai", "tensor ranking", "Vespa hybrid", "Vespa cloud" DO NOT USE FOR: other vector stores - use `vector-stores/*` siblings; ColBERT without Vespa - use `retrieval/colbert-retrieval`; classical reranking - use `rag/reranking`

Warp Rust web framework using filters. Covers routing, filters, rejection, WebSocket, and TLS. Use for composable, type-safe Rust APIs. USE WHEN: user mentions "warp", "rust filters", "composable rust api", asks about "warp filters", "warp rejection", "filter composition rust", "rust hyper warp", "warp websocket" DO NOT USE FOR: Axum projects - use `axum` instead, Actix-web projects - use `actix-web` instead, Rocket projects - use `rocket` instead, non-Rust backends

Yup schema validation library. Popular with Formik for form validation. Use when building forms with React/Formik or needing schema-based validation. USE WHEN: user mentions "Yup", "Formik validation", asks about "yup schema", "Formik setup", "form validation library" DO NOT USE FOR: Zod projects (use zod skill), NestJS/class-validator (use class-validator skill), new projects (prefer Zod)

Agent-driven RAG patterns. Self-RAG, Corrective RAG (CRAG) with web fallback, Adaptive RAG with routing classifier, ReAct with retrieval tool, multi-hop retrieval, plan-and-execute, LangGraph state machines for RAG. USE WHEN: user mentions "agentic RAG", "Self-RAG", "Corrective RAG", "CRAG", "Adaptive RAG", "multi-hop retrieval", "LangGraph RAG", "ReAct RAG", "plan and execute RAG" DO NOT USE FOR: static retrieval pipelines - use `rag-architecture`; query rewriting only - use `query-transformations`; evaluation - use `rag-evaluation`

Authentication flow validation between frontend and backend. Covers JWT, OAuth2, token refresh, and session management sync. USE WHEN: user asks about "auth integration", "JWT validation", "token refresh", "401 handling", "authentication flow", "login integration" DO NOT USE FOR: security auditing - use security skills, OAuth provider setup - use authentication skills

Axum Rust web framework by Tokio. Covers routing, handlers, extractors, middleware, and state. Use for ergonomic async Rust APIs. USE WHEN: user mentions "axum", "tokio web", "rust async api", "tower middleware", asks about "axum extractors", "axum state", "axum router", "rust websocket axum", "hyper server", "ergonomic rust api" DO NOT USE FOR: Actix-web projects - use `actix-web` instead, Rocket projects - use `rocket` instead, Warp projects - use `warp` instead, non-Rust backends

Batch RAG for high-volume ingest and bulk query scenarios. Covers OpenAI Batch API (50% discount, 24h SLA), Anthropic Message Batches API, Voyage AI and Cohere batch embeddings, ingestion-time vs query-time batching, async/Ray parallelism, parallel writes to vector DBs, and rate-limit coordination across workers. USE WHEN: user mentions "batch API", "OpenAI batch", "Anthropic batches", "bulk embedding", "Ray embeddings", "parallel ingest", "batch RAG" DO NOT USE FOR: streaming single-query latency - use `rag-production`; cost-dashboards - use `cost-allocation`; LLM gateway routing - use `llm-gateway`

ColBERT / ColBERTv2 late interaction as a first-stage retriever — not just a reranker. MaxSim scoring, PLAID index, Ragatouille for deployment, storage cost of token-level embeddings, when to pick ColBERT over dense-plus-rerank, and fine-tuning for domain. USE WHEN: user mentions "ColBERT", "ColBERTv2", "late interaction", "MaxSim", "PLAID", "Ragatouille", "token-level embeddings", "multi-vector retrieval" DO NOT USE FOR: single-vector dense retrieval - use `vector-stores/qdrant-advanced` or similar; cross-encoder reranking - use `rag/reranking`; SPLADE sparse - use `retrieval/splade-deep`; cross-encoder fine-tuning - use `retrieval/cross-encoder-training`

CI/CD for RAG quality: golden dataset fixtures, RAGAS/DeepEval in pytest, regression thresholds, GitHub Actions workflows, merge-blocking gates, weekly scheduled eval, LangSmith/Langfuse in CI. USE WHEN: user mentions "RAG CI", "eval in CI", "regression gate", "golden dataset fixture", "PR quality check", "scheduled RAG evaluation", "LangSmith CI", "Langfuse CI" DO NOT USE FOR: RAGAS metric internals - use `rag-evaluation`; ARES - use `ares-framework`; Giskard internals - use `giskard-rag`; shadow deploys - use `shadow-mode-deployment`

Per-tenant, per-feature, per-query RAG cost tracking. Covers token counting (tiktoken, Anthropic count_tokens), structured metadata logging, aggregation in BigQuery/Snowflake/ClickHouse, dashboards (Grafana, Metabase), LangSmith and Langfuse native cost reports, and budget alerts. Schema + example queries. USE WHEN: user mentions "RAG cost", "cost per tenant", "cost per query", "token counting", "chargeback", "showback", "LangSmith cost", "Langfuse cost", "budget alerts" DO NOT USE FOR: reducing cost - use `rag-production`; batch-discount strategy - use `batch-inference`; routing for cheaper models - use `llm-gateway`

Windows driver debugging end-to-end: WinDbg / WinDbg-Preview, KDNET kernel debugging setup over Ethernet/USB, Driver Verifier, Application Verifier (UMDF), bugcheck analysis with !analyze -v, common kernel commands (!process, !thread, !irp, !devstack, !wdfdevice, !wdfkd.wdfdriverinfo), live and post-mortem (kernel dumps), symbol setup, and WPP / ETW trace analysis. USE WHEN: user mentions "WinDbg", "kd", "KDNET", "kernel debugger", "Driver Verifier", "verifier.exe", "Application Verifier", "appverif", "!analyze", "!irp", "bugcheck", "0x9f DRIVER_POWER_STATE_FAILURE", "kernel dump", "MEMORY.DMP", "DebugView", "tracefmt", "tracerpt", "WPP decode" DO NOT USE FOR: Visual Studio managed/.NET debugging, web app debugging

Windows driver signing and certification: EV (Extended Validation) code-signing certificates, Microsoft Hardware Dev Center attestation signing, WHQL/HLK certification, INF Section requirements, dual-signing, kernel driver signing policy (Win10 1607+), test signing, cross-signing legacy notes, and signtool/signability validation. USE WHEN: user mentions "driver signing", "EV cert", "attestation signing", "WHQL", "HLK", "Hardware Dev Center", "Partner Center", "signtool", "test signing", "DSE bypass", "cross-signed driver", "INF compliance", "dual-signed kernel driver" DO NOT USE FOR: Authenticode signing of normal apps, certificate management unrelated to drivers, macOS / Linux signing

Echo Go web framework. Covers routing, middleware, binding, context, and WebSocket. Use for high-performance, minimalist Go APIs. USE WHEN: user mentions "echo", "labstack echo", "go echo framework", asks about "echo middleware", "echo context", "echo binding", "echo websocket", "high performance go api", "echo router" DO NOT USE FOR: Gin projects - use `gin` instead, Fiber projects - use `fiber` instead, Chi projects - use `chi` instead, non-Go backends

Embedding model selection across providers (OpenAI, Voyage, Cohere, BGE, E5, Jina, Nomic, mixedbread). Covers MTEB benchmarks, dimensions, cost, latency, multilingual support, instruction-tuning, and query-vs-document modes. USE WHEN: user mentions "which embedding model", "text-embedding-3", "voyage-3", "cohere embed", "BGE", "E5", "nomic-embed", "mxbai", "MTEB", "embedding benchmark", "embedding dimensions", "embedding cost" DO NOT USE FOR: multilingual-specific tradeoffs - use `multilingual-embeddings`; fine-tuning models - use `embedding-fine-tuning`; MRL truncation - use `matryoshka-embeddings`; vector database choice - use `vector-stores/*`

ESLint 9 flat config and Biome linter configuration. Covers typescript-eslint, rule categories, migration from legacy config, and performance optimization. USE WHEN: user mentions "ESLint 9", "Biome", "flat config", "eslint.config.mjs", asks about "typescript-eslint", "migration to ESLint 9", "linter performance", "Biome vs ESLint" DO NOT USE FOR: legacy ESLint - use `eslint` skill, Java/Spring - use SonarQube, basic linting concepts - use `quality-common`

Haystack 2.x pipeline architecture for RAG and LLM apps. Covers Components, Pipeline as DAG, document stores (InMemory, Elasticsearch, Weaviate, Pinecone, Qdrant), Embedders, Retrievers (BM25, embedding, hybrid), Generators, PromptBuilder, conditional routing, and evaluation components. USE WHEN: user mentions "Haystack", "deepset", "Haystack pipeline", "Component DAG", "DocumentStore", "BM25Retriever", "ConditionalRouter" DO NOT USE FOR: LlamaIndex specifics - use `llamaindex`; LangChain - use `langchain`; DSPy - use `dspy`

Combining sparse (BM25, SPLADE) and dense vector retrieval. Reciprocal rank fusion with formula and code, weighted score fusion, alpha tuning, and native hybrid indexes in Pinecone, Qdrant, Weaviate. USE WHEN: user mentions "hybrid search", "BM25", "sparse dense", "RRF", "reciprocal rank fusion", "SPLADE", "learned sparse", "alpha tuning", "sparse dense hybrid" DO NOT USE FOR: rewriting queries before retrieval - use `query-transformations`; reranking after retrieval - use `reranking`; vector DB setup - use `vector-databases`

Java and Spring Boot security patterns. Covers Spring Security, dependency auditing, secure coding practices, and OWASP for Java ecosystem. USE WHEN: user works with "Java", "Spring Boot", "Spring Security", asks about "Java vulnerabilities", "Maven security", "Gradle security", "Java injection", "Java authentication" DO NOT USE FOR: general OWASP concepts - use `owasp` or `owasp-top-10` instead, Node.js/Python security - use language-specific skills

Open source license compliance and SPDX standards. Covers license types, compatibility, auditing with license-checker, and SBOM generation. USE WHEN: user mentions "license", "SPDX", "GPL", "MIT", "Apache", asks about "license compatibility", "license-checker", "copyleft", "proprietary compliance", "OSI approved" DO NOT USE FOR: dependency vulnerabilities - use `supply-chain`, security scanning - use `owasp-top-10`, secrets - use `secrets-management`

LlamaIndex 0.12+ for RAG and agent applications. Covers Document/Node model, IngestionPipeline, NodeParser variants, VectorStoreIndex, query engines, sub-question decomposition, router engines, Property Graph Index, LlamaParse integration, and observability callbacks. USE WHEN: user mentions "LlamaIndex", "llama_index", "VectorStoreIndex", "IngestionPipeline", "PropertyGraphIndex", "LlamaParse", "SubQuestionQueryEngine" DO NOT USE FOR: LangChain specifics - use `langchain`; RAG architecture theory - use `rag-patterns`; DSPy compiler workflows - use `dspy`

Decision framework: when long-context (Gemini 2M, Claude 200k, GPT 128k) beats RAG, hybrid approaches (RAG narrows, long-context reads), cost-quality-latency tradeoffs, lost-in-the-middle / context rot research, needle vs synthesis tasks, prompt caching economics, concrete $ per query math. USE WHEN: user mentions "long context vs RAG", "Gemini 2M", "lost in the middle", "context rot", "when not to use RAG", "stuff the prompt", "prompt caching cost" DO NOT USE FOR: implementing RAG - use `rag-architecture`; evaluating RAG - use `rag-evaluation`; chunking decisions alone - use `chunking-strategies`

MapStruct Java mapping framework. Covers basic mappers, update mappings, nested objects, collections, custom methods, and Lombok integration. Based on castellino and gestionale-presenze production patterns. USE WHEN: user mentions "mapstruct", "@Mapper", "@Mapping", asks about "DTO mapping", "entity conversion", "@MappingTarget", "nested mapping" DO NOT USE FOR: Java language - use `java` skill instead DO NOT USE FOR: Lombok - use `lombok` skill instead DO NOT USE FOR: JPA entities - use JPA-specific skills

MongoDB Atlas Vector Search. $vectorSearch aggregation stage, index definition JSON, quantization (scalar, binary), hybrid with Atlas Search text via $rankFusion, dynamic schema benefits, sharding, Atlas Triggers for auto-embedding. USE WHEN: user mentions "Atlas Vector Search", "$vectorSearch", "MongoDB vector", "$rankFusion", "Atlas Triggers embedding", "MongoDB HNSW" DO NOT USE FOR: self-hosted vector stores - use `vector-stores/qdrant-advanced`, `vector-stores/milvus`; MongoDB without Atlas (self-hosted) - vector search is an Atlas-only feature

Retrieval over images, PDFs with figures and tables, audio, and video. Vision-language embeddings (CLIP, SigLIP, ImageBind, VoyageAI multimodal, BGE-M3), vision-model ingestion (Claude Vision, GPT-4o Vision), table-aware retrieval, Whisper-based audio RAG, keyframe + transcript video RAG. USE WHEN: user mentions "multimodal RAG", "image search", "PDF with figures", "table extraction", "audio RAG", "video RAG", "CLIP", "SigLIP", "VoyageAI multimodal", "BGE-M3", "Claude Vision RAG", "GPT-4o Vision" DO NOT USE FOR: pure text RAG - use `rag-patterns`; graph reasoning - use `graph-rag`; grounding/citation enforcement - use `rag-guardrails`

Amazon OpenSearch k-NN plugin. nmslib vs faiss vs lucene engine, HNSW vs IVF + PQ, radial search, neural search with remote embedding, hybrid search (BM25 + kNN with weighted sum), OpenSearch 2.13+ features, memory estimation, sizing. Full index mapping. USE WHEN: user mentions "OpenSearch", "OpenSearch k-NN", "knn_vector", "neural search", "nmslib", "faiss engine OpenSearch", "AWS vector search" DO NOT USE FOR: other vector stores - use `vector-stores/*` siblings; Elasticsearch (proprietary) - OpenSearch diverged from Elasticsearch 7.10; classical BM25 tuning - use `retrieval/bm25-tuning`

Advanced Qdrant features. Payload indexing, scalar/binary/product quantization, named vectors (multi-vector), sparse vectors, multi-tenancy, snapshots, sharding, gRPC vs HTTP, pre- vs post-filtering, and oversampling/rescoring. USE WHEN: user mentions "Qdrant", "named vectors", "sparse vectors", "scalar quantization qdrant", "qdrant filtering", "qdrant sharding", "qdrant multi-tenant" DO NOT USE FOR: basic Qdrant upsert/query - use `ai-integration/vector-databases`; other vector DBs - use other `vector-stores/*`; quantization theory - use `vector-quantization`

Pre-retrieval query rewriting techniques. HyDE, multi-query, step-back, RAG-fusion with RRF, sub-query decomposition, query routing, and expansion. Full Python code per technique with LangChain and native Anthropic SDK. USE WHEN: user mentions "HyDE", "hypothetical document", "multi-query", "step-back prompting", "RAG-fusion", "query rewriting", "query decomposition", "query routing" DO NOT USE FOR: post-retrieval reranking - use `reranking`; sparse+dense fusion on retrieved docs - use `hybrid-search`; agent loops - use `agentic-rag`

Caching strategies across the RAG stack. Semantic caching with GPTCache and LangChain, Redis-based embedding-similarity cache, cache key design, TTL/invalidation, partial caching (cache retrieval only), provider-native prompt caching (Anthropic, OpenAI), and hierarchical L1/L2 caches. USE WHEN: user mentions "semantic cache", "GPTCache", "LLM cache", "prompt caching", "Redis vector cache", "cache invalidation for RAG", "reduce LLM cost", "latency reduction LLM" DO NOT USE FOR: retrieval accuracy - use `rag-patterns`; groundedness checks - use `rag-guardrails`; incremental indexing - use `rag-production`

Evaluating RAG systems. RAGAS, DeepEval, TruLens, custom LLM-as-judge, golden datasets, synthetic test generation, A/B testing, retrieval-only metrics (Hit@K, MRR, NDCG), and answer quality metrics. USE WHEN: user mentions "RAGAS", "RAG evaluation", "faithfulness", "answer relevancy", "context precision", "hit rate", "MRR", "NDCG", "LLM as judge", "golden dataset", "synthetic data" DO NOT USE FOR: runtime telemetry only - use standard monitoring; chunking decisions - use `chunking-strategies`; overall design - use `rag-architecture`

Trust-and-safety layer for RAG. Hallucination detection (LLM self-check, NLI entailment, TRUE metric), groundedness scoring, forced-citation prompting, out-of-scope refusal, NeMo Guardrails, Guardrails AI, Pydantic structured outputs with source verification, LLM-as-judge validators. USE WHEN: user mentions "hallucination detection", "groundedness", "citation enforcement", "NeMo Guardrails", "Guardrails AI", "LLM-as-judge", "refusal", "faithfulness", "TRUE metric", "answer verification" DO NOT USE FOR: prompt-injection via retrieved docs - use `rag-security`; retrieval quality metrics (recall/precision) - use `rag-observability`; caching of verified answers - use `rag-caching`

Tracing, evaluation, and alerting for RAG systems. LangSmith, Langfuse, Arize Phoenix, Comet Opik, OpenTelemetry GenAI conventions. What to log (query, chunks+scores, rerank scores, answer, citations), retrieval debugging workflows, alerts for empty/low-score retrieval. USE WHEN: user mentions "LangSmith", "Langfuse", "Phoenix", "Opik", "OpenTelemetry LLM", "LLM tracing", "retrieval debugging", "RAG metrics", "RAG dashboard", "RAG alerting" DO NOT USE FOR: hallucination validators - use `rag-guardrails`; caching layer metrics - use `rag-caching`; security audit logs - use `rag-security`

Redis Stack vector search. FT.CREATE with VECTOR field, HNSW vs FLAT, hybrid with other field types, RedisVL Python library, Redis Enterprise scaling, pipelines for bulk ingest, TTL for ephemeral vectors (cache use case), Redis JSON + vectors pattern. USE WHEN: user mentions "Redis vector", "Redis Stack", "FT.CREATE", "RedisVL", "Redis HNSW", "Redis JSON vector", "semantic cache Redis" DO NOT USE FOR: other vector stores - use `vector-stores/*` siblings; embedding caching theory - use `rag/caching-retrieval`

Shadow and canary deployment of RAG pipeline changes: dual-execute new + old, offline LLM-judge comparison, gradual traffic ramp, auto-rollback guardrails, multi-armed bandits, per-component feature flags (retriever, reranker, generator). USE WHEN: user mentions "shadow mode", "canary deployment", "dual execute", "shadow traffic RAG", "feature flags RAG", "multi-armed bandit", "auto rollback", "LaunchDarkly RAG", "Unleash RAG" DO NOT USE FOR: CI pre-merge eval - use `continuous-evaluation`; offline eval only - use `rag-evaluation`; drift detection - use `drift-detection`

Smoke testing patterns for verifying backend implementations end-to-end. Covers build verification, service health checks, authentication flows, test data setup, and HTTP endpoint validation. USE WHEN: user mentions "smoke test", "verify implementation", "test endpoints", "check if it works", "end-to-end verification", "does it actually run" DO NOT USE FOR: Unit test writing - use `vitest` or `junit`; E2E browser tests - use `playwright`; Load testing - use `performance-expert`; Static API validation - use `integration-validator-expert`

Spring Data JDBC for simple, lightweight database access without JPA complexity. Covers aggregates, repositories, custom queries, and DDD patterns. USE WHEN: user mentions "spring data jdbc", "simple database access", "no JPA", "aggregate roots", "DDD with JDBC", "lightweight ORM", "@MappedCollection" DO NOT USE FOR: JPA/Hibernate features - use `spring-data-jpa` instead, reactive database - use `spring-r2dbc` instead, NoSQL - use respective skills

Spring LDAP for LDAP/Active Directory integration and authentication. Covers LdapTemplate, @Entry mapping, LdapRepository, and Spring Security integration. USE WHEN: user mentions "LDAP", "Active Directory", "AD authentication", "LdapTemplate", "@Entry", "directory services", "enterprise authentication" DO NOT USE FOR: OAuth2/OIDC - use `spring-security` or `oauth2` skill, simple auth - use Spring Security with database

Spring Security JWT authentication for Spring Boot 3. Covers SecurityFilterChain, JWT filter, authentication, authorization, CORS, rate limiting, and security headers. USE WHEN: user mentions "security", "JWT", "authentication", "authorization", asks about "login", "token", "CORS", "roles", "permissions", "@PreAuthorize" DO NOT USE FOR: Spring Boot basics (use `spring-boot`), OAuth2 details (use OAuth2 skill if available), general encryption (use security-expert)

Vector compression techniques: scalar quantization (int8, int4), binary quantization (bit vectors + Hamming), product quantization (PQ), optimized PQ, residual quantization. Covers rescoring workflow, storage math, accuracy tradeoffs, and Faiss/Qdrant/pgvector implementations. USE WHEN: user mentions "quantization", "int8 embeddings", "binary embeddings", "product quantization", "PQ", "OPQ", "residual quantization", "Hamming distance", "compress embeddings", "rescoring" DO NOT USE FOR: Matryoshka dimension truncation - use `matryoshka-embeddings`; ANN algorithm choice - use `ann-algorithms`

ARES (Stanford) automated RAG evaluation: synthetic query generation, fine-tuned classifiers for faithfulness and relevance, prediction-powered inference (PPI) for unbiased estimates from a small gold set. Compared to RAGAS. USE WHEN: user mentions "ARES", "Stanford ARES", "prediction-powered inference", "PPI for RAG", "ares-ai", "fine-tuned RAG judges", "automated RAG eval with small gold set" DO NOT USE FOR: LLM-as-judge only workflow - use `rag-evaluation` (RAGAS); Giskard tooling - use `giskard-rag`; CI integration patterns - use `continuous-evaluation`

Bun JavaScript runtime. Fast all-in-one toolkit with bundler, test runner, package manager. Use when working with Bun projects or considering Node.js alternatives. USE WHEN: user mentions "bun", "bun test", "bun build", asks about "Bun.serve", "bun install", "SQLite in Bun", "bunx", "performance comparison" DO NOT USE FOR: Node.js runtime - use `nodejs` skill instead DO NOT USE FOR: Hono/Elysia frameworks - use framework-specific skills DO NOT USE FOR: Language syntax - use `typescript` or `javascript` skills

C++ memory and concurrency safety: AddressSanitizer (ASan), UndefinedBehaviorSanitizer (UBSan), ThreadSanitizer (TSan), MemorySanitizer (MSan), MSVC `/sdl` and `/guard:cf`, CERT C++ secure coding rules, integer-overflow safe arithmetic, and Control Flow Integrity. USE WHEN: user mentions "ASan", "AddressSanitizer", "UBSan", "TSan", "MSan", "use-after-free", "buffer overflow", "undefined behavior", "data race", "CERT C++", "secure C++", "/sdl", "/GS", "ASLR", "DEP", "stack canary" DO NOT USE FOR: Web/app security (use `owasp`), .NET security (`dotnet-security`), cryptography APIs (use `cryptography`)

Detecting distributional drift in production embeddings. Metrics (KL, PSI, MMD, cosine-to-centroid), alerting thresholds, query-vs-corpus divergence, root-cause analysis, and embedding observability tooling (Arize, Fiddler). USE WHEN: user mentions "embedding drift", "distribution shift", "PSI", "KL divergence", "MMD", "embedding monitoring", "embedding observability", "model staleness", "production embedding quality" DO NOT USE FOR: evaluation of answer quality - use `rag-evaluation`; retraining embeddings - use `embedding-fine-tuning`; general model drift on tabular features - use an ML monitoring skill

DSPy 2.5+ for programming (not prompting) LMs. Covers Signatures, Modules (Predict, ChainOfThought, ReAct, Retrieve), compilers/optimizers (BootstrapFewShot, MIPROv2, BootstrapFinetune), retrievers, evaluation, and production deployment of compiled programs. USE WHEN: user mentions "DSPy", "dspy.Module", "dspy.Signature", "MIPROv2", "BootstrapFewShot", "ChainOfThought optimization", "program > prompt" DO NOT USE FOR: manual prompt templates - use `langchain`; LlamaIndex query engines - use `llamaindex`; generic RAG architecture - use `rag-patterns`

Elasticsearch 8.15+ vector search. dense_vector with HNSW, kNN search, hybrid BM25 + kNN with RRF, reranker integration, ELSER learned-sparse, sparse_vector, int8 scalar quantization, and index performance tuning. USE WHEN: user mentions "elasticsearch vector", "dense_vector", "kNN search", "ELSER", "sparse_vector elastic", "elasticsearch RRF", "elasticsearch hybrid search" DO NOT USE FOR: classic full-text only - use Elasticsearch search skills; other vector DBs - use other `vector-stores/*`

Fine-tuning sentence embedding models with sentence-transformers. Contrastive objectives (MultipleNegativesRankingLoss, TripletLoss, CoSENT), domain adaptation, synthetic query generation (GPL), hard negative mining, and downstream retrieval evaluation. USE WHEN: user mentions "fine-tune embeddings", "domain adaptation", "sentence-transformers training", "contrastive loss", "hard negatives", "GPL", "synthetic queries", "MNR loss", "triplet loss" DO NOT USE FOR: picking a pretrained model - use `embedding-models`; MRL truncation - use `matryoshka-embeddings`; reranker training - out of scope

Standardized error response handling between frontend and backend. Covers error structure, status codes, and consistent error handling. USE WHEN: user asks about "error handling", "API errors", "error response format", "status codes", "error messages", "validation errors" DO NOT USE FOR: logging - use logging skills, exception handling - use language-specific skills

Deep hard-negative mining for dense retrievers: in-batch vs mined, BM25 / dense / cross-encoder mining, TAS-B, MarginMSE distillation, curriculum schedules, memory banks, false-negative filtering. USE WHEN: user mentions "hard negatives", "negative mining", "TAS-B", "MarginMSE", "distillation from cross-encoder", "curriculum negatives", "memory bank", "MoCo", "false negative filter" DO NOT USE FOR: basic fine-tuning setup - use `embedding-fine-tuning`; reranker training - out of scope; picking a pretrained model - use `embedding-models`

JVM performance profiling with Java Flight Recorder (JFR), jcmd, and GC analysis. Use for identifying bottlenecks and memory issues. USE WHEN: user mentions "Java profiling", "JFR", "JVM performance", asks about "Java Flight Recorder", "jcmd", "heap dump", "GC tuning", "thread dump", "Java memory leak" DO NOT USE FOR: Node.js/Python profiling - use respective skills instead

JSDoc and TSDoc documentation standards. Covers syntax, tags, TypeScript integration, documentation generation, and API documentation best practices. USE WHEN: user mentions "TSDoc", "TypeScript documentation", "API documentation", asks about "documenting TypeScript", "@param", "@returns", "TypeDoc", "API Extractor", "documentation tags" DO NOT USE FOR: JavaScript-only JSDoc - use `jsdoc` skill for plain JavaScript

Milvus 2.4+ distributed vector database. Architecture (etcd + MinIO + Pulsar), partitions, index types (IVF_FLAT, HNSW, DiskANN, GPU_IVF_PQ), scalar field indexing, hybrid search (BM25 + vector), Zilliz Cloud managed, dynamic schema, upserts, time travel, pymilvus v2 client. USE WHEN: user mentions "Milvus", "pymilvus", "Zilliz", "DiskANN", "IVF_FLAT", "Milvus partition", "GPU vector index" DO NOT USE FOR: other vector DBs - use `vector-stores/*` siblings; general vector index theory - use `rag/vector-quantization`

OpenAPI contract validation between frontend and backend. Covers type sync, endpoint validation, and contract-first development. USE WHEN: user asks about "OpenAPI validation", "API contract", "swagger sync", "frontend backend types", "API mismatch", "contract testing" DO NOT USE FOR: GraphQL - use `graphql-contract` skill, authentication - use `auth-flow-validation` skill

Advanced pgvector 0.7+ on PostgreSQL. HNSW vs IVFFlat tuning, halfvec / bit / sparsevec types, binary quantization, hybrid search with pg_trgm + tsvector, parallel index builds, partitioning for multi-tenancy, replication caveats, and Supabase-specific patterns. USE WHEN: user mentions "pgvector", "postgres vector", "HNSW in postgres", "IVFFlat", "halfvec", "supabase vector", "pgvector hybrid search", "pgvector tuning" DO NOT USE FOR: non-Postgres vector DBs - use other `vector-stores/*` skills; embedding model choice - use `embedding-models`

pytest-django integration testing specialist. Covers all fixtures (db, transactional_db, client, rf, settings, mailoutbox, django_user_model), @pytest.mark.django_db options, DRF APIClient, factory_boy integration, async views, signals, management commands, and multi-database testing. USE WHEN: user mentions "pytest-django", "django test", "@pytest.mark.django_db", asks about "django client fixture", "DRF APIClient", "django signals test", "management command test", "django async view test". DO NOT USE FOR: Non-Django Python tests - use `pytest` or `python-integration`; FastAPI - use `fastapi-testing`; Pure container setup - use `testcontainers-python`

RAG system architecture and design decisions. Covers naive vs advanced vs agentic RAG, decision trees for RAG vs fine-tuning vs long context, production topology, latency budgets, and component sequencing. USE WHEN: user mentions "RAG architecture", "RAG design", "naive RAG", "advanced RAG", "agentic RAG", "RAG vs fine-tuning", "RAG vs long context", "production RAG" DO NOT USE FOR: chunking details - use `chunking-strategies`; query rewriting - use `query-transformations`; retrieval algorithms - use `advanced-retrieval`; evaluation - use `rag-evaluation`; agent loops - use `agentic-rag`

React Hook Form with Zod validation. Covers form setup, validation schemas, field components, error handling, and submission with shadcn/ui integration. USE WHEN: user mentions "React Hook Form", "RHF", "Zod validation", "shadcn/ui forms", "zodResolver", "FormField", asks about "form validation with Zod", "shadcn form components" DO NOT USE FOR: React 19 Server Actions - use `react-19` skill instead, basic form handling - use `react-forms` skill instead, controlled/uncontrolled patterns - use `react-forms` skill instead

SPLADE / SPLADE++ learned sparse retrieval in depth. How SPLADE differs from BM25 (learned term expansion), FLOPS regularization, indexing in Qdrant sparse vectors and Elasticsearch, hybrid with dense, efficiency tradeoffs, and when SPLADE beats BM25. USE WHEN: user mentions "SPLADE", "SPLADE++", "learned sparse", "neural sparse", "FLOPS regularization", "sparse vector retrieval", "naver/splade" DO NOT USE FOR: classical BM25 tuning - use `retrieval/bm25-tuning`; dense retrieval - use `vector-stores/*`; hybrid fusion math - use `rag/hybrid-search`

Netflix Eureka for service discovery in Spring Cloud microservices. Covers Eureka Server, Eureka Client, registration, and health checks. USE WHEN: user mentions "eureka", "service discovery", "service registry", "@EnableEurekaServer", "@EnableDiscoveryClient", "eureka dashboard" DO NOT USE FOR: Kubernetes service discovery - use native K8s, Consul - use Consul-specific patterns

Spring Cloud Gateway for API routing, filtering, and load balancing. Covers route predicates, filters, rate limiting, and circuit breaker integration. USE WHEN: user mentions "spring cloud gateway", "API gateway", "route predicates", "gateway filters", "rate limiting gateway", "load balancing gateway" DO NOT USE FOR: simple reverse proxy - use nginx, Zuul (deprecated) - use Spring Cloud Gateway instead

Spring Data JPA for database access in Spring Boot applications. Covers repositories, entities, relationships, queries, pagination, and auditing. Based on production patterns from castellino and gestionale-presenze projects. USE WHEN: user mentions "JPA", "repository", "entity", "database queries", asks about "pagination", "Criteria API", "Specifications", "@Query", "relationships" DO NOT USE FOR: Spring Boot basics (use `spring-boot`), Spring Security (use `spring-security`), MongoDB (use `mongodb-expert`), SQL-only tasks (use `sql-expert`)

Spring Data MongoDB for Java/Spring Boot applications. Covers repositories, MongoTemplate, aggregations, and document mapping. USE WHEN: user mentions "spring data mongodb", "MongoTemplate", "MongoRepository", "@Document", "Spring Boot MongoDB", "aggregation pipeline Java" DO NOT USE FOR: raw MongoDB driver - use `mongodb` instead, relational databases - use `spring-data-jpa` or `spring-data-jdbc` instead

Bean Validation (JSR-380) with Spring Boot. Covers request validation, custom validators, validation groups, and error handling. USE WHEN: user mentions "spring validation", "@Valid", "@NotNull", "@NotBlank", "bean validation", "request validation", "custom validator", "validation groups", "ConstraintValidator", "@Pattern" DO NOT USE FOR: business rule validation - use service layer, security validation - use `spring-security` skill

Kernel-Mode Driver Framework (KMDF), the Microsoft-recommended way to write Windows kernel-mode drivers. Covers DriverEntry, EvtDeviceAdd, IRPs and IOCTLs, I/O queues, PnP and Power state machines, IRQL discipline, memory pools, WPP tracing, SAL annotations, and Driver Verifier. USE WHEN: user mentions "KMDF", "WDF kernel", "Windows kernel driver", "DriverEntry", "WdfDriverCreate", "EvtDeviceAdd", "IRP", "IOCTL", "DISPATCH_LEVEL", "PASSIVE_LEVEL", "NTSTATUS", "PoolTag", "WdfRequestComplete" DO NOT USE FOR: UMDF v2 (use `wdf-umdf`), classic WDM-only drivers, file-system filters (FltMgr is a separate framework)

Webpack module bundler. Covers configuration, loaders, plugins, optimization. Use when working with Webpack-based projects or migrating from Webpack. USE WHEN: user mentions "Webpack", "webpack.config", "webpack loaders", "webpack plugins", asks about "Webpack configuration", "bundle optimization" DO NOT USE FOR: Vite projects (use vite skill), esbuild (use esbuild skill), new projects (prefer Vite), Parcel

Approximate Nearest Neighbor algorithm theory and selection. HNSW, IVF, IVF+PQ, ScaNN, DiskANN, LSH. Covers parameter effects (m, ef, nlist, nprobe), recall/latency/memory tradeoffs, exhaustive vs approximate, and when to pick which index. USE WHEN: user mentions "ANN", "HNSW", "IVF", "nearest neighbor", "ScaNN", "DiskANN", "LSH", "recall vs latency", "index type choice", "FAISS index" DO NOT USE FOR: specific DB configuration - use the respective `vector-stores/*-advanced` skill; quantization details - use `vector-quantization`

Pinecone's Canopy framework — a pre-packaged RAG stack on top of Pinecone. Covers Canopy CLI, REST server (OpenAI-compatible), chunking, embedding, retrieval, chat-history management, and customization hooks (custom chunkers/records). USE WHEN: user mentions "Canopy", "Pinecone Canopy", "canopy start", "canopy new", "Pinecone out-of-the-box RAG" DO NOT USE FOR: custom pipelines on non-Pinecone stores - use `rag-architecture`; Postgres-based stacks - use `r2r`; LangChain/LlamaIndex abstractions - use `langgraph-rag`/`llamaindex`

AST-aware chunking for source code RAG. Covers tree-sitter, LangChain Language splitter, LlamaIndex CodeSplitter, preserving function/class boundaries, keeping docstrings with implementations, and code-specific embeddings (voyage-code-2, CodeBERT). USE WHEN: user mentions "code chunking", "tree-sitter", "code splitter", "AST chunking", "code embedding", "voyage-code-2", "code RAG", "repo Q&A" DO NOT USE FOR: plain prose chunking - use `rag-patterns`; PDF extraction - use `pdf-extraction`

Design system architecture and design tokens. Covers W3C Design Token spec, atomic design methodology, component documentation, theming, dark/light mode, and modern CSS custom property patterns. USE WHEN: user mentions "design system", "design tokens", "component library", "theming", "dark mode", "light mode", "CSS variables", "atomic design", "color palette", "token architecture", "shadcn theming", "brand tokens" DO NOT USE FOR: specific component APIs (use styling/shadcn-ui or styling/radix-ui), animation timing (use ux/interaction-design), WCAG audit (use accessibility/wcag)

esbuild extremely fast JavaScript bundler. Used internally by Vite. Use for simple bundling, library builds, or understanding Vite internals. USE WHEN: user mentions "esbuild", "fast bundler", asks about "esbuild config", "library bundling", "esbuild API" DO NOT USE FOR: Complex apps (use Vite), Webpack projects (use webpack skill), need extensive plugins (use Rollup/Webpack)

factory_boy test data generation specialist. Covers Factory, DjangoModelFactory, SQLAlchemyModelFactory, all field declarations (Faker, LazyAttribute, Sequence, SubFactory, RelatedFactory, post_generation, Trait, Maybe, Dict, List), batch creation, pytest integration, and Celery task testing patterns. USE WHEN: user mentions "factory_boy", "test factory", "DjangoModelFactory", "SQLAlchemyModelFactory", asks about "test data generation", "factory traits", "SubFactory", "factory fixtures". DO NOT USE FOR: pytest internals - use `pytest`; Django setup - use `pytest-django`; Hypothesis property testing - use `pytest` with Hypothesis

Flyway database migrations for Spring Boot applications. Covers migration scripts, versioning, callbacks, and rollback strategies. Based on production patterns from castellino and gestionale-presenze projects. USE WHEN: user mentions "flyway", "database versioning", "Java migrations", "Spring Boot migrations", "V1__", "migration checksum" DO NOT USE FOR: general migration strategies - use `migrations` instead, Liquibase - use specific tool skill, Prisma/TypeORM - use respective skills

Giskard RAGET (RAG Evaluation Toolkit): automatic testset generation (simple / complex / distracting / conversational), component-level scoring (retriever / generator / rewriter), hallucination and bias tests, CI integration. Compared to RAGAS and DeepEval. USE WHEN: user mentions "Giskard", "RAGET", "Giskard RAG toolkit", "automatic testset generation", "component-level RAG scoring", "hallucination test Giskard" DO NOT USE FOR: general RAGAS usage - use `rag-evaluation`; Stanford ARES - use `ares-framework`; CI/CD wiring - use `continuous-evaluation`

Go code quality with golangci-lint, staticcheck, and go vet. Covers linting, formatting, and best practices. USE WHEN: user works with "Go", "Golang", "Gin", "Fiber", "Echo", asks about "golangci-lint", "staticcheck", "go vet", "Go linting", "Go best practices" DO NOT USE FOR: SonarQube - use `sonarqube` skill, testing - use Go test skills, security - use `go-security` skill

JavaScript language fundamentals. Covers ES6+, modules, async patterns. Use when working with plain JS projects without TypeScript. USE WHEN: user mentions "javascript", "ES6", "async/await", "promises", "modules", asks about "destructuring", "spread operator", "arrow functions", "ESM vs CommonJS" DO NOT USE FOR: TypeScript projects - use `typescript` skill instead DO NOT USE FOR: Node.js runtime - use `nodejs` skill instead DO NOT USE FOR: Browser APIs - use framework-specific skills

LanceDB columnar vector database. Arrow-native storage, versioning and time-travel, merge-on-read, full-text + vector hybrid, pandas/polars integration, object-storage backing, Rust-based performance, embedding function registration, IVF_PQ and HNSW indexes. USE WHEN: user mentions "LanceDB", "Lance format", "Arrow vector store", "embedded vector DB", "pylance", "lance time travel" DO NOT USE FOR: managed vector DBs - use `vector-stores/pinecone-advanced`, `vector-stores/mongodb-atlas-vector`; distributed Milvus - use `vector-stores/milvus`

Late chunking (Jina method): embed the full document with a long-context encoder, then mean-pool token embeddings per chunk. Preserves cross-chunk context lost by traditional chunk-then-embed pipelines. USE WHEN: user mentions "late chunking", "Jina late chunking", "long-context embeddings", "cross-chunk context", "mean pool token embeddings", "jina-embeddings-v3", "token-level pooling for retrieval" DO NOT USE FOR: choosing chunk sizes - use `chunking-strategies`; picking an embedder - use `embedding-models`; fine-tuning - use `embedding-fine-tuning`

Project Lombok for reducing Java boilerplate. Covers annotations for getters, setters, constructors, builders, logging, and more. Based on production patterns from castellino and gestionale-presenze projects. USE WHEN: user mentions "lombok", "@Data", "@Builder", "@Slf4j", asks about "boilerplate reduction", "getters/setters", "@RequiredArgsConstructor", "@Value" DO NOT USE FOR: Java language features - use `java` skill instead DO NOT USE FOR: MapStruct integration - use `mapstruct` skill DO NOT USE FOR: Spring annotations - use `backend-spring-boot` skill

WebSocket integration for NestJS using Socket.IO. Covers gateways, rooms, authentication guards, exception filters, and Redis adapter for scaling. USE WHEN: user mentions "NestJS WebSocket", "NestJS gateway", "Socket.IO NestJS", "real-time NestJS", "NestJS chat", "WebSocket authentication NestJS", asks about "how to implement WebSocket in NestJS" DO NOT USE FOR: React WebSocket hooks - use `react-websocket` instead, Spring WebSocket - use `spring-websocket` instead, raw Node.js WebSocket - use `nodejs` skill

Node.js performance profiling with V8 CPU profiler, heap analysis, and perf_hooks. Use for identifying bottlenecks and memory leaks. USE WHEN: user mentions "Node.js performance", "profiling", "memory leak", asks about "V8 profiler", "heap snapshot", "CPU profile", "perf_hooks", "event loop lag", "Node.js optimization" DO NOT USE FOR: Java/Python profiling - use respective skills instead

Optical Character Recognition for scanned documents and images. Covers Tesseract, EasyOCR, PaddleOCR, AWS Textract, Azure Document Intelligence, Google Document AI, and Anthropic Claude vision. Preprocessing (deskew, denoise, binarize). USE WHEN: user mentions "OCR", "scanned PDF", "Tesseract", "Textract", "EasyOCR", "PaddleOCR", "Document AI", "Claude vision OCR", "image to text" DO NOT USE FOR: born-digital PDFs with selectable text - use `pdf-extraction`; table-only extraction - use `table-extraction`