neo4j-contrib/neo4j-skills
24 skills26 installs1.9k starsGitHub
Install
npx skills add https://github.com/neo4j-contrib/neo4j-skillsSkills in this repo
1Neo4j Cypher Skillneo4j-cypher-skill is a deep Cypher companion for Neo4j 2025.x and 2026.x: it helps agents write correct reads and writes, structure subqueries and batch paths, and catch performance landmines before they hit production. Coverage spans LOAD CSV, schema inspection (SHOW INDEXES/CONSTRAINTS, db.schema.visualization, apoc.meta.schema when APOC exists), EXPLAIN/PROFILE validation, vector and fulltext search paths, and modern language features like dynamic labels, type predicates, and OrNull casting. Defaults assume 2025.01-safe syntax with explicit annotations for newer releases. Solo builders use it when implementing graph backends, refactoring hot queries, or teaching an agent your indexing strategy. It deliberately excludes driver migrations and pure DBA tasks—those live in sibling Neo4j skills. Expect intermediate-to-advanced graph modeling familiarity; the skill shines when you already know your domain but need precise, version-aware Cypher.2installs2Neo4j Modeling Skillneo4j-modeling-skill is a Databases agent skill from neo4j-contrib that steers coding assistants through designing, reviewing, and refactoring Neo4j graph models. Solo builders shipping knowledge graphs, recommendations, or identity-heavy SaaS use it when the hardest risk is wrong topology—supernodes, vague labels, or properties that should be relationships—before implementation sprawls. The skill targets greenfield domain mapping, reviews of existing graphs, and relational or document migrations into labeled property graphs with constraints and indexes. It deliberately stops at modeling: query writing belongs to neo4j-cypher-skill, Spring mapping to neo4j-spring-data-skill, GraphQL to neo4j-graphql-skill, and ingest to neo4j-import-skill. Prism lists it as draft/WIP; still valuable for agent-led design sessions with WebFetch and Bash available per skill metadata.2installs3Neo4j Agent Memory Skillneo4j-agent-memory-skill teaches solo builders how to implement durable, queryable agent memory using the neo4j-agent-memory Python package and either self-hosted Neo4j 5.x/2025.x or the Neo4j Agent Memory Service at memory.neo4jlabs.com. It covers short-term message threads, long-term POLE+O entities (Person, Object, Location, Event, Organisation), and reasoning traces so agents can recall context across sessions without relying on a flat vector-only pile. You learn graph schema inspection with Cypher, NAMS API key configuration, rate limits, and an MCP server surface for Claude Code–style tool ecosystems. The skill contrasts graph-native memory with vector-only approaches and points to sibling skills for vector indexes, GraphRAG, and general MCP setup. Install via pip and npx skills add from neo4j-contrib/neo4j-skills. Intermediate builders shipping custom agents get the most value when they need explainable, structured recall—not just embedding search.1installs4Neo4j Aura Agent Skillneo4j-aura-agent-skill packages a solo-builder workflow for Neo4j Aura Agents—GraphRAG assistants backed by AuraDB. It targets indie developers who already have graph data in Aura and want repeatable agent setup instead of one-off console clicking. The skill walks schema discovery (including vector index detection), maps real graph shape to the three first-class tools, and drives agent lifecycle through authenticated REST calls with prompts and exposure modes. Install when you are wiring RAG or analytics chat into a product and need agents that respect your ontology rather than generic retrieval. Prerequisites are explicit: enabled Generative AI and Aura Agent in the org, project admin rights, client credentials, and org/project IDs from the console. Use uv-synced Python helpers to list agents, create from config, and smoke-test NL queries before you expose MCP or public REST endpoints to end users.1installs5Neo4j Aura Graph Analytics SkillNeo4j Aura Graph Analytics Skill teaches agents how to use Neo4j’s serverless GDS environment: ephemeral sessions billed per minute, isolated from embedded plugins. Solo builders adding recommendations, fraud rings, or community detection to a product get step-by-step guidance for Aura API credentials, choosing memory tiers, creating and reconnecting sessions, projecting graphs remotely from AuraDB or external stores, running mutate stream and write modes, polling async jobs, and streaming properties back into applications. It explicitly separates concerns from embedded GDS and from raw Cypher skills, reducing mixed camelCase and snake_case mistakes in the Python client. Advanced but narrow—ideal when your architecture already points at Aura Business Critical or VDC and graphdatascience 1.15 or newer.1installs6Neo4j Aura Provisioning Skillneo4j-aura-provisioning-skill is a draft agent skill for solo and indie builders who need Neo4j Aura online without guessing CLI flags or API flows. It walks through provisioning managed graph instances, managing lifecycle operations, and handling secrets in pipelines so agents do not leak credentials in chat logs. The same package explains Aura Agent, Neo4j’s no-code console agent that loops through interpret → plan tools → execute read-only Cypher → respond, which suits quick retrieval assistants and prototypes before you commit to a custom LangGraph stack. Prerequisites are explicit: enable Generative AI assistance and Aura Agent in org settings, and ensure tool authentication is on for the project (default for newer orgs). Use it when you are standing up Aura for an AI feature, graph RAG, or knowledge graph MVP and want repeatable steps your coding agent can follow end to end.1installs7Neo4j Document Import Skillneo4j-document-import-skill is a build-phase agent skill for solo builders who want unstructured sources—PDFs, articles, API dumps—modeled as a Neo4j knowledge graph instead of flat vectors only. It walks your coding agent through practical import paths: chunk documents with strategy-aware splitters, run LLM entity and relationship extraction (including SimpleKGPipeline), use the LLM Graph Builder for lighter-touch setup, bulk-load with apoc.load.json, or plug in LangChain and LlamaIndex loaders. The extended reference deepens chunking tradeoffs and entity resolution when real corpora collide. Catalog status is draft/WIP, so treat outputs as starting points and validate against your schema and Neo4j version. Best when you already have Neo4j in the stack and need repeatable import playbooks the agent can follow without guessing splitter APIs.1installs8Neo4j Driver Dotnet SkillNeo4j Driver Dotnet Skill is a reference-style agent skill that keeps solo .NET builders from guessing Neo4j.Driver v6 APIs when shipping graph-backed features. It documents installation, singleton driver lifecycle with VerifyConnectivityAsync, dependency-injection registration, and when to choose ExecutableQuery versus ExecuteReadAsync and ExecuteWriteAsync managed transactions. Coverage extends through IResultCursor consumption, safe record access, Cypher-to-.NET type mapping including temporal types and ElementId caveats, UNWIND batching with dictionaries or anonymous types, and Preview-level AsObject mapping. Error handling emphasizes exception hierarchy and rollback-safe patterns. The skill deliberately points Cypher authoring to neo4j-cypher-skill and upgrades to neo4j-migration-skill so this package stays focused on connection and execution correctness. Use it while building APIs, workers, or internal tools that persist relationships in Neo4j; skip it if you are on another language driver or not using Neo4j at all.1installs9Neo4j Driver Go SkillNeo4j Driver Go Skill packages procedural knowledge for building backend Go services on Neo4j Go Driver v6. Indie developers shipping APIs or internal tools use it when they need correct connection handling, query execution style, and error semantics without reading the entire driver manual. The skill centers neo4j.ExecuteQuery as the default happy path, then documents when managed read/write transactions or explicit BeginTransaction better fit streaming or cross-function work. It maps Go types to Cypher safely, propagates context deadlines, and explains retry behavior around connectivity failures. Batch ingestion via UNWIND and bookmark management for causal consistency are included for production-shaped code. It deliberately excludes Cypher authoring and v5→v6 migration, pointing to sibling neo4j-skills for those concerns. Advanced pool sizing and repository wrappers live in referenced markdown for when defaults are not enough.1installs10Neo4j Driver Java SkillNeo4j-driver-java-skill teaches agents how to write production-grade Java and Kotlin against the official Neo4j Java Driver v6. Solo builders and small teams reach for it when a product needs graph relationships—recommendations, permissions, knowledge graphs—without guessing transaction boundaries or leaking sessions. The skill spans dependency setup, managed and explicit transactions, async and reactive entry points, and practical mapping rules such as null-safety with asString and containsKey. It deliberately excludes Cypher authoring, version migrations, and Spring Data Neo4j, pointing to sibling skills for those concerns so the agent does not mix ORM patterns with the low-level driver. Compatibility is stated as Driver v6, Java 17+, and Kotlin 1.9+. Use it during backend implementation when stack traces involve ServiceUnavailableException, TransientException, or incorrect parameter typing in UNWIND batches.1installs11Neo4j Driver Python SkillNeo4j Driver Python Skill teaches solo builders how to connect Python services to Neo4j with the official neo4j package (v6.x, Python 3.10+). It is for indie SaaS and agent backends that store relationships—not flat rows—and need reliable sessions, routing, retries, and async FastAPI integration without drowning in driver edge cases. Coverage spans installation, URI schemes, execute_query as the default API, managed transactions for retry-safe writes, implicit transactions for bulk LOAD CSV flows, AsyncGraphDatabase patterns, structured error handling, Record/JSON serialization pitfalls, Python–Cypher type mapping including temporal and graph objects, UNWIND list-of-dict batching, pool tuning, and bookmarks. It deliberately defers Cypher authoring to neo4j-cypher-skill and GraphRAG to neo4j-graphrag-skill. On Prism, shelf it under Build/backend while treating driver correctness as multi-phase: you revisit it when shipping load tests and when operating production outages.1installs12Neo4j Gds Skillneo4j-gds-skill teaches agents how to operate Neo4j’s Graph Data Science plugin through the graphdatascience Python client or Cypher on Aura Pro, self-managed, local, or offline DBMS instances. Solo builders use it when recommendation systems, community detection, centrality, or embedding pipelines must run inside Neo4j rather than in a separate notebook-only stack. The readme spells compatibility bands for GDS client v1.21, embedded plugin on Neo4j 5.x, Python 3.10–3.14, and the Neo4j driver, and it deliberately defers Cypher authoring, driver bootstrap, vector index DDL, and Aura Graph Analytics Sessions to sibling neo4j-* skills. Intermediate complexity reflects choosing projection strategy, execution mode, and memory-safe runs before mutating or writing graph properties. Install path is pip install graphdatascience; you are expected to already have GDS installed on the target database.1installs13Neo4j Genai Plugin SkillNeo4j GenAI plugin skill teaches agents to use Neo4j’s native ai.text.* functions so solo builders can run embeddings, text completion, structured JSON outputs, and chat without bolting a separate orchestration service onto every retrieval query. It fits builders shipping knowledge-heavy agents or SaaS features where the graph is the system of record: configure providers, discover available models, chunk by tokens, and chain vector search with traversals and LLM steps inside Cypher. The skill is explicit about version gates—2025.12+, CYPHER 25, Aura GenAI enabled—and points vector index mechanics to a sibling skill. That makes it a focused integration reference, not a full Python GraphRAG pipeline guide. Install when you are implementing GraphRAG or entity-aware copilots on Neo4j; skip if you only need index DDL or offline batch ETL in another language. Expect agents to need API credentials and careful review of queries that touch production graphs.1installs14Neo4j Getting Started SkillNeo4j Getting-Started Skill orchestrates a solo builder or autonomous agent from no database to a working Neo4j-backed application through eight sequential stages, each driven by its own reference markdown so context stays lean. Stages cover prerequisites, project context, Aura provisioning, graph data modeling, synthetic or real data load, exploration, query exercises, and a minimal app or notebook build. It expects Bash, file I/O, and Neo4j MCP tools for Cypher and schema work, with optional data-model validation and visualization MCPs. Choose it when starting a greenfield graph feature—not when you only need Cypher snippets, driver upgrades, or day-two CLI operations, which the skill defers to neo4j-cypher, migration, and CLI skills. Time budgets are documented for autonomous versus human-in-the-loop runs.1installs15Neo4j Graphql Skillneo4j-graphql-skill guides solo builders through building a GraphQL API backed by Neo4j using the Neo4j GraphQL Library—currently documented for v7 with v5 LTS paths. It covers constructor setup, index and constraint assertion, rich type definitions, relationship directions, custom @cypher resolvers, JWT/JWKS security on fields, auto-generated CRUD, OGM for trusted server-side access, subscriptions, and federation hooks with Apollo. Invoke it when you are writing typeDefs, securing fields, or wiring Neo4j to Apollo Server, and reach for neo4j-cypher-skill when you need Cypher outside GraphQL resolvers. The README marks the skill as draft/WIP; treat examples as living and verify against your installed @neo4j/graphql major version. Intermediate complexity assumes you already run a Neo4j instance and understand basic graph modeling.1installs16Neo4j Graphrag Skillneo4j-graphrag-skill teaches solo builders to assemble production-style GraphRAG on Neo4j with the neo4j-graphrag Python package (v1.16.0+). It covers when to use each retriever—from pure vector and hybrid search through graph-augmented VectorCypher and HybridCypher flows, natural-language Text2Cypher, and ToolsRetriever for routing across multiple backends including Weaviate, Pinecone, and Qdrant. You learn how to craft retrieval_query fragments, pass query_params and filters, connect an LLM and prompt into the GraphRAG class, and tune quality when answers feel shallow or off-topic. The skill documents embedders, index creation, token tracking, newer Cypher SEARCH patterns, and hooks for LangChain or LlamaIndex. It deliberately does not ingest documents into a KG, run GDS graph algorithms, or manage long-horizon agent memory—those belong to neo4j-document-import-skill, neo4j-vector-index-skill, neo4j-gds-skill, and neo4j-agent-memory-skill respectively.1installs17Neo4j Import SkillNeo4j-import-skill guides coding agents through importing structured CSV data into Neo4j with the right tool for scale and environment. Solo builders shipping knowledge graphs, recommendation backends, or analytics on Aura benefit from explicit branching: Neo4j Data Importer for sub-million-row, low-Cypher workflows with drag-and-drop files, versus LOAD CSV batched with CALL IN TRANSACTIONS for larger sets, plus pointers to neo4j-admin import for offline bulk. The skill spells access paths in Aura, standalone importer URLs, and constraints such as string-only properties in the GUI importer. It is phase-specific backend work—invoke when CSVs and a graph model are ready, not during marketing or SEO. Marked draft upstream; pair with your own validation on production-sized files.1installs18Neo4j Kafka SkillNeo4j Kafka Skill is an agent skill for solo builders and small teams who need Neo4j graphs in motion with Apache Kafka or Confluent Cloud. It documents how to install and operate the Neo4j Connector for Kafka on self-managed Connect (Confluent Hub or direct JAR) or use the managed sink in Confluent Cloud without local JAR plumbing. Coverage spans sink configuration strategies—from Cypher and pattern mapping to CDC and CUD flows—plus exactly-once semantics and dead-letter handling when records fail validation. On the source side it explains CDC-driven streaming where the edition supports it and query-based polling elsewhere, including the native db.cdc.query API and cursor-loop patterns in Python and Java. Schema Registry integration for Avro and JSON Schema is included so agents do not guess wire formats. The skill defers Cypher authoring, bulk file import, and GDS algorithms to sibling Neo4j skills, keeping this package focused on event-driven graph sync. Expect intermediate data-engineering familiarity with Connect workers, topics, and Neo4j credentials.1installs19Neo4j Nvl Skillneo4j-nvl-skill is an agent skill for solo builders shipping graph-aware SaaS or internal tools who need Neo4j Visualization Library (NVL) wired correctly the first time. It walks through @neo4j-nvl/base for nodes, relationships, rendering options, and export, @neo4j-nvl/interaction-handlers for zoom, pan, selection, and keyboard UX, and @neo4j-nvl/react wrappers for React 19 apps. You learn when to pick Canvas2D versus WebGL2, how to choose layouts from forceDirected through d3Force, and how to connect live Cypher results via nvlResultTransformer instead of hand-mapping records. The skill deliberately excludes Needle’s GraphVisualization defaults, Python/Jupyter ports, raw Cypher authoring, driver lifecycle, and GDS—pointing you to sibling Neo4j skills for those concerns. Use it while building dashboards, agent memory graphs, or analytics UIs on Neo4j 1.1+ in modern browsers.1installs20Neo4j Query Tuning Skillneo4j-query-tuning-skill is an agent skill that walks solo builders and small teams through Neo4j Cypher performance diagnosis end to end. It teaches when to use EXPLAIN versus PROFILE, how to read plans bottom-up, and which operators signal missing indexes, bad cardinality, or expensive graph traversals. The skill pairs operator-level fix strategies with planner hints and runtime selection guidance for slotted, pipelined, and parallel execution, including Enterprise-only capabilities where noted. It also covers operational tooling—SHOW QUERIES, SHOW TRANSACTIONS, transaction termination, and db.stats-style retrieval—so you can fix slow queries without guessing. Reference files deepen coverage of plan operators and stats/monitoring. Install via agentskills.io for Claude Code-style agents when your app depends on Neo4j and latency or cost spikes trace back to Cypher.1installs21Neo4j Security SkillNeo4j Security Skill teaches agents how to manage Neo4j security programmatically through Cypher and configuration patterns: users, roles, privilege grants, inspection, and Enterprise-only controls. Solo builders shipping knowledge graphs, recommendation backends, or internal tools on Neo4j need this when default admin access is no longer acceptable and RBAC must be expressed as repeatable commands. The skill documents CREATE and ALTER USER flows, role grants, DENY semantics, property-level access per label, sub-graph restrictions, and ABAC rules tied to OIDC claims, plus SHOW PRIVILEGES patterns suitable for export and audit. Community edition covers basic RBAC; property-level, sub-graph, ABAC, LDAP, and SSO scenarios require Enterprise. It deliberately excludes application query authoring, neo4j-admin cluster operations, and driver session management, deferring to neo4j-cypher-skill, neo4j-cli-tools-skill, and driver-specific skills. Install via the neo4j-contrib skills package when hardening a graph deployment or automating least-privilege rollouts alongside infra changes.1installs22Neo4j Snowflake Graph Analytics SkillNeo4j Snowflake Graph Analytics Skill teaches agents to use Neo4j Graph Analytics for Snowflake—a Native Application that exposes graph algorithms as SQL procedures on warehouse tables. Solo builders and small data teams install it when they need PageRank, Louvain, Dijkstra, node similarity, and related workloads without exporting data to a separate graph cluster. The skill walks through marketplace setup, privilege grants, compute pool sizing, projection definitions, and the repeatable project-compute-write workflow. It also helps pick algorithms for fraud detection, recommendations, and entity resolution, chain multi-step jobs, and fix projection or permission failures. Expect intermediate Snowflake and SQL fluency; outputs are runnable SQL and configuration patterns aligned with Neo4j’s current Snowflake documentation.1installs23Neo4j Spark SkillNeo4j Spark Skill teaches coding agents how to move graph data between Apache Spark and Neo4j using the official Neo4j Connector for Apache Spark. Solo builders and small data teams use it when standing up batch sync from a lakehouse into a knowledge graph, or exporting graph slices for ML feature stores on Databricks, EMR, or standalone Spark. Coverage includes SparkSession dependency wiring, mutually exclusive read options for label, Cypher, and relationship scans, and write semantics including MERGE via node.keys. You get operational knobs for partitions and batch size plus Databricks-specific install and secrets guidance. The skill explicitly defers Cypher authoring, the Python bolt driver, GDS, and Spring Data to sibling Neo4j skills so agents do not mix incompatible stacks. Pair it during build when your product needs scheduled graph hydration rather than interactive Cypher-only apps.1installs24Neo4j Vector Index Skillneo4j-vector-index-skill walks solo builders through Neo4j-native vector search so graph-backed apps can retrieve by embedding similarity instead of only explicit relationships. It covers index creation with correct dimensions and similarity functions aligned to model training loss, operational checks until indexes are ONLINE, and ingestion paths from Python batch UNWIND loops to in-database ai.text.embed and embedBatch on supported versions. Query guidance spans the newer SEARCH clause on Neo4j 2026.01+ and procedure fallbacks for earlier releases, plus patterns that combine vector hits with graph traversal for hybrid RAG. The skill also addresses chunking before ingest, using vectors from GDS algorithms, and common failure modes such as dimension mismatch, offline indexes, and null provider returns. It explicitly defers broader GenAI plugin coverage to a sibling skill. Builders on Claude Code or Cursor use it when wiring knowledge graphs, recommendation flows, or agent memory layers on Neo4j.1installs