redpanda-data

Generate Codex hooks.json and AGENTS.md from Claude Code hooks. Maps supported hooks directly, including Edit|Write apply_patch aliases, and keeps Stop batch checks only as fallback. Use when setting up Codex compatibility or dual-agent support.

Question changed code as liability and remove unjustified surface area. Use before commit, push, PR, merge, or when a change feels overbuilt or low-value.

Reviews diffs for over-engineering only. Use when reviewing, desloping, or asking what to delete, inline, shrink, or replace with stdlib/native code.

Resilience Review checks resilience. Use when edge cases, errors, fallback, async/data, state, or polish matter.

Reviews a diff since a fixed point across Standards, Spec, and risk gates. Use when reviewing a branch, PR, WIP, or "review since X".

parallel executor. Use /swarm.

Runs a release-blocking cold PR audit across quality, frontend, resilience, visual, security, tests, perf, and steelman. Use when very important PR, high-stakes PR, thermo nuclear review, or no-stones-unturned audit requested.

Ultra-compressed reply mode. Use when user asks be brief or /caveman.

Teaches how to build streaming data pipelines with Redpanda Connect (formerly Benthos) — declarative YAML config, the input/pipeline/output model, Bloblang mapping language, error handling, and canonical pipeline patterns. Use when: building or running a Redpanda Connect pipeline; wiring an input to an output with processors; writing Bloblang mappings or mutations; choosing connectors (kafka, redpanda, http_server, file, generate, sql_*, aws_*, gcp_*); running pipelines with rpk connect run or redpanda-connect; configuring buffers, caches, metrics, rate limits, or the global redpanda block; writing config.yaml for redpanda-connect; asking what Connect components exist and how to discover them; error handling with fallback outputs or catch processors; batching records before writing; understanding the difference between mapping and mutation processors; choosing between the bloblang and mapping processor names; connecting to Kafka or Redpanda with SASL/TLS from Connect; migrating Kafka clusters to Redpanda with

Provision and manage Redpanda Cloud BYOC (Bring Your Own Cloud) clusters via the Control Plane API (https://api.redpanda.com) and the rpk cloud byoc agent plugin. Covers OAuth2 client-credentials auth, creating Network resources per cloud provider (AWS VPC/subnets/IAM, GCP network/project/bucket, Azure VNet/subnets/identities), creating TYPE_BYOC clusters (network_id, throughput_tier, cloud_provider_tags, customer_managed_resources), polling long-running Operations until READY, and running the Terraform-backed rpk cloud byoc apply/destroy/validate flow with --redpanda-id. Use when: creating BYOC clusters in AWS/GCP/Azure via the public API; provisioning or tearing down BYOC networks; wiring customer-managed IAM roles/buckets/subnets into a Redpanda cluster; setting up private connectivity (AWS PrivateLink, GCP Private Service Connect, Azure Private Link, VPC/network peering via NetworkPeeringService, and AWS Transit Gateway centralized egress); registering cross-account AWS access (CloudProviderAccessService,

Configures and operates an Oxla cluster: YAML config (default_config.yml), OXLA__ environment-variable overrides, node ports and roles, leader election, storage backends (local/S3/GCS/Azure), memory limits, access control, TLS, OIDC, and the HTTP-based ConnectRPC admin service (port 9090) with its LoggingService (GetLogLevel/SetLogLevel). Also covers the Prometheus metrics endpoint (port 8080) and Docker/Compose deployment patterns. Covers Oxla's lakehouse and streaming differentiators: Apache Iceberg REST catalogs (CREATE ICEBERG CATALOG, oauth2/basic/aws_sigv4 auth, feature_flags.allow_iceberg_queries), transparent Redpanda/Kafka integration (CREATE REDPANDA CATALOG, topic-backed tables, schema_lookup_policy/error_handling_policy), object-storage connections (CREATE STORAGE TYPE S3/GCS/ABS), plus security: OIDC/JWT auth, SCRAM passwords, centralized access control, and AES-256-GCM secret encryption (OXLA_ENCRYPTION_KEY). Use when: configuring an Oxla cluster; setting ports or node names; enabling TLS or OID

Query external data from Oxla — Kafka topics via catalogs, Apache Iceberg tables, and S3/GCS/Azure parquet/ORC files — alongside native Oxla tables. Use when: querying Kafka topics with CREATE KAFKA CATALOG or CREATE REDPANDA CATALOG; reading Apache Iceberg tables using the catalog=>path.table syntax; loading or exporting parquet/ORC files from S3/GCS/Azure (COPY FROM/TO); joining external data with native tables; or inspecting external metadata via system.kafka_connections, system.kafka_sources, system.iceberg_catalogs, and system.iceberg_tables. Also covers the Redpanda Enterprise features that produce the data Oxla reads — Iceberg Topics (redpanda.iceberg.mode and the redpanda.iceberg.* topic properties; iceberg_enabled and iceberg_rest_catalog_* cluster properties), Tiered Storage (cloud_storage_enabled), and Server-Side Schema ID Validation (enable_schema_id_validation, redpanda.value.schema.id.validation) — all of which require a Redpanda Enterprise license. Trigger phrases: "query Kafka topic from Oxla

Write and run SQL against Oxla, a PostgreSQL-wire-compatible distributed columnar analytical database. Covers connecting via psql or any PostgreSQL driver (port 5432, password "oxla" by default), the full supported data-type set (INT/LONG/INT16/INT32/FLOAT/DOUBLE/CHAR/VARCHAR/STRING/TEXT/DATE/TIME/ TIMESTAMP/TIMESTAMPTZ/INTERVAL/BOOL/JSON/JSONB/BYTEA/NUMERIC/ARRAY/GEOMETRY/ GEOGRAPHY/POINT), DDL (CREATE/DROP TABLE, CREATE TABLE AS SELECT, CREATE/DROP VIEW, CREATE/DROP SCHEMA, TRUNCATE, CREATE ROLE/GRANT/REVOKE), DML (SELECT/INSERT/UPDATE/DELETE, SELECT INTO table or file), data loading (COPY FROM/TO with CSV/Parquet/ORC formats, S3 credentials), aggregate and window functions (SUM/AVG/COUNT/MIN/MAX/percentile_disc/percentile_cont/mode, ROW_NUMBER/RANK/DENSE_RANK/LAG/LEAD/FIRST_VALUE/LAST_VALUE/NTH_VALUE/NTILE/ CUME_DIST), CTEs (WITH), UNION/INTERSECT/EXCEPT (with optional ALL), PREPARE/EXECUTE, and analytic query patterns (GROUP BY, ORDER BY, LIMIT/OFFSET, multi-table joins, star-schema aggregations). Also co

Collects local and remote Redpanda diagnostics bundles and gathers broker process info using the `rpk debug` command group. Use when: collecting a debug bundle for Redpanda support or self-triage, generating a remote (cluster-wide) bundle via the Admin API, gathering local broker process or diagnostic info from the CLI, using rpk debug bundle or rpk debug remote-bundle, troubleshooting a Redpanda cluster and needing to capture logs/metrics/profiles, passing a debug bundle to the support team, or running rpk debug on Linux or Kubernetes. Also covers triaging enterprise (license-gated) features from a bundle — Tiered Storage, Cloud Topics, Iceberg Topics, Continuous Data Balancing, Shadow Linking (cross-cluster DR), Remote Read Replicas, Audit Logging, RBAC/GBAC, OIDC/OAuthBearer/ Kerberos auth, FIPS mode, Server-side Schema ID Validation, Schema Registry Authorization, and Leadership Pinning, plus checking license status and license violations. Applies to self-managed deployments only (Linux hosts and Kubernet

Manage Redpanda topics and produce/consume records from the CLI using the `rpk topic` command group. Covers creating topics with custom partition counts, replication factors, and configs; listing and describing topics (summary, configs, partitions); altering topic configs (set/delete/append/ subtract); adding partitions; trimming/deleting records; analyzing throughput and batch size; describing tiered-storage status; and deleting topics. Also covers `rpk topic produce` and `rpk topic consume` in depth including the --format percent-escape syntax, keys, headers, compression, acks, tombstones, schema-registry encoding/decoding, and all --offset forms. Also covers Redpanda enterprise topic properties (Enterprise license required): Tiered Storage (redpanda.remote.read/write/delete/recovery, redpanda.storage.mode, retention.local.target.*), Cloud Topics (redpanda.cloud_topic.enabled / redpanda.storage.mode=cloud), Iceberg Topics (redpanda.iceberg.mode/delete/partition.spec/target.lag.ms/ invalid.record.action), Re

Builds, deploys, and manages Redpanda Data Transforms — WebAssembly (Wasm) functions that run inside the broker to transform records in-flight from an input topic to one or more output topics. Covers the full lifecycle: enabling the feature, initializing a project, writing transform logic in Go (TinyGo), Rust, JavaScript, or TypeScript, building the Wasm binary, deploying with rpk transform deploy, listing and inspecting running transforms, viewing logs, pausing/resuming, and deleting transforms. Use when: writing or deploying a Redpanda data transform, in-broker Wasm transform, wiring an input topic to an output topic, filtering or transforming Kafka records inside the broker, using rpk transform init / build / deploy / list / logs / pause / resume / delete, enabling data_transforms_enabled cluster config, using the Go transform SDK, Rust transform SDK, or JavaScript/TypeScript transform SDK, from-offset reprocessing, or setting environment variables on a transform. Also covers wiring transform output topics

Client-side React performance optimization patterns.

Connect Query patterns for API calls. Use when working with mutations, queries, or data fetching.

TypeScript, React, and JavaScript best practices enforced by Ultracite/Biome.

Write and run Playwright E2E tests for Redpanda Console using testcontainers. Analyzes test failures, adds missing testids, and improves test stability. Use when user requests E2E tests, Playwright tests, integration tests, test failures, missing testids, or mentions 'test workflow', 'browser testing', 'end-to-end', or 'testcontainers'.

Refactor legacy forms to use modern Redpanda UI Registry Field components with react-hook-form and Zod validation. Use when user requests: (1) Form refactoring or modernization, (2) Converting Chakra UI or @redpanda-data/ui forms, (3) Updating forms to use Field components, (4) Migrating from legacy form patterns, (5) Implementing forms with react-hook-form and Zod validation.

Manage client and server state with Zustand stores and React Query patterns.

Build UI with Redpanda Registry components, Tailwind v4, and accessibility best practices.

Write and maintain tests with Vitest v4 dual configuration, mock utilities, and Zustand store testing patterns.

Code review a pull request for Redpanda Connect, checking Go patterns, tests, component architecture, and commit policy

Audit every consumer of the schema.Common metadata format (the format produced by schema_registry_decode's store_schema_metadata, the parquet_decode processor, and CDC sources) for type-coverage drift and value-coercion gaps. Run this whenever a new component starts consuming schema.Common, when a new schema.CommonType variant is added upstream in benthos, or as a periodic maintenance check.

This skill should be used when users need to create or debug Bloblang transformation scripts. Trigger when users ask about transforming data, mapping fields, parsing JSON/CSV/XML, converting timestamps, filtering arrays, or mention "bloblang", "blobl", "mapping processor", or describe any data transformation need like "convert this to that" or "transform my JSON".

This skill should be used when users need to discover Redpanda Connect components for their streaming pipelines. Trigger when users ask about finding inputs, outputs, processors, or other components, or when they mention specific technologies like "kafka consumer", "postgres output", "http server", or ask "which component should I use for X".

This skill should be used when users need to create or fix Redpanda Connect pipeline configurations. Trigger when users mention "config", "pipeline", "YAML", "create a config", "fix my config", "validate my pipeline", or describe a streaming pipeline need like "read from Kafka and write to S3".

Use when adding or modifying a Redpanda admin v2 ConnectRPC endpoint — authoring the protobuf under proto/redpanda/core/admin/v2/, wiring it into the Bazel build, implementing the C++ service handler under src/v/redpanda/admin/services/, registering it on the admin server, and regenerating Python bindings for ducktape integration tests. Covers both new RPCs on an existing service and brand-new services.

"Create a new backport branch with prefix `ai-backport-` by cherry-picking all commits from a GitHub PR onto a target release branch or commit. Use when the user wants to backport a PR, create a backport branch, cherry-pick PR commits to a release branch, or mentions backporting changes from dev/main to a version branch (e.g. v25.1.x, v25.2.x). Requires two arguments: TARGET (the release branch or commit hash to backport onto) and PR_NUM (the GitHub PR number to backport)."

Improve code coverage for a Bazel C++ test target or source file

Use when scaffolding a brand-new resource or datasource in the terraform-provider-redpanda repo. Triggers on "add resource", "new datasource", "scaffold redpanda <thing>", or any work that creates a fresh redpanda/resources/<name>/ directory. Covers schema yaml, schemagen wiring, model generation, hand-written CRUD glue, provider registration, unit tests, acceptance tests, examples, and docs end-to-end. For modifying an existing resource use extend-redpanda-resource instead.

Use when adding or modifying fields/attributes on an EXISTING resource or datasource in terraform-provider-redpanda. Triggers on "add field X to redpanda_<name>", "extend the topic resource", "new attribute on redpanda_cluster datasource". For brand-new resources or datasources (no existing package), use add-redpanda-resource instead.

Use when diagnosing a bug in terraform-provider-redpanda — whether a user reports a runtime failure ("expected empty plan but resource has planned action(s)", "WARN ... removing from state", "Provider produced inconsistent result after apply") OR a proactive coverage audit finds a function whose purpose is uncovered at one or more test tiers. Walks the mock/fake parity investigation, layered-defense design, red-test-first sequencing, and tier-by-tier regression-guard placement. Covers the specific "mock-loop passes but live run fails" bug class that has been the dominant failure mode in this repo. For new feature work use add-redpanda-resource or extend-redpanda-resource instead.

Guidance for adding or updating LLM models in this repo's provider catalogs (providers/*/models.go). Use when registering a new model, a new model snapshot, or new region/profile variants, especially on Bedrock. Pairs with the "Adding New Models" pricing rules in CLAUDE.md.

Author Postgres migrations when a persisted proto changes, manage the protoc-gen-go-jet generation pipeline, wire new resources, and diagnose drift. The `protoc-gen-go-jet diff` subcommand either copies ddl.sql verbatim (first-ever migration) or emits an LLM brief (the caller authors the ALTER SQL beneath it); `protoc-gen-go-jet check` re-runs the whole chain and fails if it diverges from ddl.sql. Trigger when a proto with `(gojet.v1.table)` changes, a new persisted resource is added, or `protoc-gen-go-jet check` reports drift.