# Build Streaming Applications with Agent Skills Confluent agent skills are packaged, AI-guided workflows that help AI coding assistants build streaming applications step by step. With agent skills, you get code generation, decision-making support, and guidance on Schema Registry integration, security configuration, and error handling. Use Confluent agent skills to build Apache Kafka® producers and consumers, Kafka Streams applications, and CDC pipelines. ## How agent skills work When you ask your AI coding assistant for help with a Confluent task, your assistant activates the relevant skill automatically based on your request. You receive a guided workflow that walks you through the task with code tailored to your specific use case and environment. Confluent agent skills follow the [Agent Skills Specification](https://agentskills.io/specification), an open standard for packaging AI-guided workflows. With Confluent agent skills, you get best practices built in at development time. For example, when you scaffold a Kafka producer, the skill automatically includes Schema Registry serialization, proper security configuration, and error handling, rather than leaving these as optional add-ons. ## Supported AI agents You can use Confluent agent skills with the following AI coding assistants: - [Claude Code](https://docs.anthropic.com/en/docs/claude-code) - [Cursor](https://www.cursor.com/) - [GitHub Copilot](https://github.com/features/copilot) You can also use any AI coding assistant that supports the [Agent Skills Specification](https://agentskills.io/specification). ## Available skills The following table lists the Confluent agent skills available for installation. For more detail on each skill, see the [agent skills GitHub repository](https://github.com/confluentinc/agent-skills). | Skill | What you can do | Prerequisites | |--------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| | confluent-cloud-cdc-tableflow | Set up end-to-end Change Data Capture (CDC) pipelines on
Confluent Cloud using Debezium source connectors, Apache Flink® for
transformation, and Tableflow for data lake integration. | Confluent Cloud account,
[open-source MCP server](https://github.com/confluentinc/mcp-confluent)
or Confluent CLI | | confluent-skill-creator | Build and test new Confluent product skills end to
end. Covers scoping, drafting, spec-compliance validation,
and end-to-end tests against a real Confluent
environment. | Python 3.9+, the `confluent-kafka` library (with Avro,
JSON Schema, and Protobuf extras), access to a Confluent Cloud,
Confluent Platform, local Docker, or WarpStream environment for testing | | confluent-skill-reviewer | Audit a Confluent agent skill against the Agent Skills
Specification, Confluent conventions, and eval coverage
requirements. Get a structured report of blocking issues,
warnings, and nits. | A skill directory or PR in the
[agent-skills](https://github.com/confluentinc/agent-skills)
repository | | developing-kafka-java-client | Scaffold or extend Java Kafka producer and consumer
projects with Schema Registry serialization (Avro, JSON Schema, or
Protobuf). Targets Confluent Cloud, local Docker, or WarpStream
using Maven or Gradle. | Java 17+, Maven or Gradle, Docker (for local development) | | developing-kafka-python-client | Scaffold Python Kafka producer and consumer projects using the
`confluent-kafka-python` library with Schema Registry serialization
(Avro, JSON Schema, or Protobuf). You can choose between
asynchronous and synchronous modes. | Python, pip, Docker (for local development) | | flink-udf | Build and deploy Flink user-defined functions (UDFs) in
Java: scalar UDFs, user-defined table functions (UDTFs),
and process table functions (PTFs). Invoke them from
Flink SQL or the Table API on Confluent Cloud or local Docker. | Java 17+, Maven or Gradle, Confluent CLI or Docker (for
local development) | | kafka-schema-registry | Scan your projects for Kafka applications, extract schemas from
data models, tag PII fields, generate Terraform for Schema Registry
registration, and produce a migration report with rollout
ordering. | [Terraform](https://www.terraform.io/) (to apply generated
configurations) | | kafka-streams-programming | Architect, build, and debug Kafka Streams applications. You get
help with topology design, pattern selection (joins, windows,
aggregations), code generation, and troubleshooting for
production issues. | Java 17+, Gradle or Maven, Docker (for local development) | | msk-migration | Assess and plan a migration from Amazon Managed
Streaming for Apache Kafka (AWS MSK) to Confluent Cloud. Includes an
environment readiness audit and a technical migration
plan covering cluster sizing, networking, authentication,
and switchover approach. | An AWS MSK environment to assess;
[KCP](https://github.com/confluentinc/kcp) (optional,
for automated environment scans) | ## Install agent skills Claude Code has a native plugin system for installing skills. For Cursor, GitHub Copilot, and other compatible agents, use the Skills CLI. ### Install in Claude Code To add the Confluent agent skills marketplace entry, run the following command: ```bash /plugin marketplace add confluentinc/agent-skills ``` To install the skills plugin, run the following command: ```bash /plugin install streaming-skills-plugin@confluent-agent-skills ``` ### Install with the Skills CLI For other AI coding assistants, use the Skills CLI to install Confluent agent skills. The Skills CLI runs through the Node Package Execute (`npx`) tool and requires [Node.js 18 or later](https://nodejs.org/). ```bash npx skills add confluentinc/agent-skills ``` This command downloads the skills and configures them for any compatible AI coding assistant in your project. ## Use agent skills After you install skills, open a project where you want to build streaming applications and ask your AI coding assistant for help. Your assistant activates the relevant skill automatically based on your request. You can also invoke a skill directly by typing `/` followed by the skill name, for example `/kafka-streams-programming`. You do not need Confluent Cloud credentials unless your workflow connects to a cluster. Example prompts: - “Create a Kafka producer that sends user profile events to Confluent Cloud” - “Scan this project for Kafka applications and generate Schema Registry Terraform” - “Build a Kafka Streams application that joins order and customer events” - “Set up a CDC pipeline from PostgreSQL to an Iceberg table on Confluent Cloud” ## Uninstall agent skills To remove Confluent agent skills from Claude Code, run: ```bash /plugin uninstall streaming-skills-plugin@confluent-agent-skills ``` To remove skills you installed with the Skills CLI, run: ```bash npx skills remove confluentinc/agent-skills ``` ## Limitations - Skill outputs are generated by AI models, which are nondeterministic. Review and test all generated code before you deploy to production. - You can use skills to guide development workflows, but they do not deploy infrastructure or make changes to Confluent Cloud resources directly. ## Feedback and contributions If you have feedback on existing skills or want to request new skills for specific use cases, you can [open an issue](https://github.com/confluentinc/agent-skills/issues) in the agent skills repository. To contribute a skill, review the [contributing guide](https://github.com/confluentinc/agent-skills/blob/main/CONTRIBUTING.md). ## Related content - [Agent skills GitHub repository](https://github.com/confluentinc/agent-skills) (GitHub) - [Agent Skills Specification](https://agentskills.io/specification) - [Use AI Tools with Confluent Cloud](overview.md#ai-tools-overview) - [Use the Managed MCP Server](managed-mcp-server.md#ai-tools-managed-mcp) - [Use the Open-Source MCP Server](open-source-mcp-server.md#ai-tools-oss-mcp)