Streaming Debugging

Diagnose a self-managed Redpanda broker or cluster with debug bundles, the metrics endpoints, logs, CPU profiling, and partition and Raft health checks, and look up the details in the Redpanda docs. Use when a broker is unhealthy, crashing, or out of disk, when partitions are under-replicated or leaderless, when consumer lag keeps growing, when collecting a debug bundle for Redpanda support with rpk debug bundle or rpk debug remote-bundle, or when an Enterprise feature such as Continuous Data Balancing, Tiered Storage, or Shadow Linking misbehaves. For raw Admin API endpoints, see /redpanda:st

9 stars 4 forksSoftware Development Plugin marketplace

Install

Claude Code plugin
/plugin marketplace add redpanda-data/skills

The repo is a Claude Code plugin marketplace. Run /plugin afterwards to pick the plugin that includes this skill.

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add redpanda-data/skills --skill streaming-debugging
Manual install
git clone --depth 1 https://github.com/redpanda-data/skills.git /tmp/skills
mkdir -p ~/.claude/skills
cp -r /tmp/skills/skills/streaming-debugging ~/.claude/skills/streaming-debugging

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# Redpanda Streaming debugging

Debugging a self-managed cluster means reading its health, metrics, logs, and
debug bundles, then fixing the cause. This skill keeps only the safety rule
below. Look up everything else in the docs (last section).

## How to work

**Confirm a broker is gone for good before you force recovery or decommission
it.** Forcing recovery of partitions that lost majority
(`rpk cluster partitions unsafe-recover`, or the Admin API
`force_recover_from_nodes`) can lose every record the surviving replicas
didn't have. Decommissioning removes the broker from the cluster for good and
moves all its data. A broker that is only down rejoins and catches up when it
restarts. So when brokers are down, say what restarting them would restore and
what each forced step would cost, then ask the user whether the brokers can
come back. A `--dry` run is fine without asking.

**To stop partitions moving, take the balancer off `continuous` first, then
cancel the moves.** With `partition_autobalancing_mode=continuous`, cancelling
the current movements is not enough, because the balancer schedules new ones on
its next round. Set the mode to `off` (or `node_add`), then run
`rpk cluster partitions move-cancel`, and remind the user to set the mode back
afterward.

## Use the docs for everything else

Look up commands, flags, endpoints, metric names, properties, and
troubleshooting procedures in the Redpanda docs instead of from memory:
Read the full skill on GitHub

Details

Skill name
streaming-debugging
Path
skills/streaming-debugging/SKILL.md
License
Apache-2.0
Repo updated
16 November 2026

FAQ

What does the Streaming Debugging skill do?

Diagnose a self-managed Redpanda broker or cluster with debug bundles, the metrics endpoints, logs, CPU profiling, and partition and Raft health checks, and look up the details in the Redpanda docs. Use when a broker is unhealthy, crashing, or out of disk, when partitions are under-replicated or leaderless, when consumer lag keeps growing, when collecting a debug bundle for Redpanda support with rpk debug bundle or rpk debug remote-bundle, or when an Enterprise feature such as Continuous Data Balancing, Tiered Storage, or Shadow Linking misbehaves. For raw Admin API endpoints, see /redpanda:st

How do I install the Streaming Debugging skill?

Claude Code plugin: /plugin marketplace add redpanda-data/skills. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Streaming Debugging popular?

It lives in redpanda-data/skills, which has 9 GitHub stars and 4 forks.

Related skills

Adp

9

redpanda-data/skills

Redpanda's Agentic Data Plane: governance infrastructure for building, running, and governing AI agents and MCP servers behind a credential-injecting AI Gateway for LLM providers, operated through the `rpk ai` CLI and the ADP UI at ai.redpanda.com. Use when installing `rpk ai`, signing in to an environment, checking setup with `rpk ai doctor`, managing managed or self-managed AI agents and their triggers with `rpk ai agent` and `rpk ai trigger`, configuring remote or managed MCP servers with `rpk ai mcp-server` (including code mode and user-delegated OAuth), setting up LLM providers and models

AI, LLMs & Agents

Adp Governance

9

redpanda-data/skills

Govern what Agentic Data Plane agents may do, spend, and expose, through the `rpk ai` CLI and the ADP UI at ai.redpanda.com, and look up governance features in the Redpanda docs. Use when writing or debugging a Cedar access policy with `rpk ai policy`, giving a new agent access, capping agent spend with `rpk ai budget`, analyzing cost, attaching a guardrail to a provider, containing or releasing an agent with the kill switch, storing or deleting a credential with `rpk ai secret`, minting an API key, masking MCP tool data with data policies, or setting up inbound or outbound OAuth. For signing

AI, LLMs & Agents

Adp Observability

9

redpanda-data/skills

Read what Agentic Data Plane agents did and why, through traces, conversation transcripts, the audit log, and the monitoring views in the ADP UI at ai.redpanda.com and the `rpk ai` CLI. Use when finding which call made an agent run slow, failed, or expensive with `rpk ai trace`, reading or filtering a conversation with `rpk ai agent transcript`, finding conversations with a failed tool call, investigating a denied or masked call with `rpk ai auditlog`, working out on whose behalf an agent acted, or reading the Agent network or Home views. For budgets, cost, and access policies, see `/redpanda:

DevOps & Cloud

Backend Go Expert

9

ydnikolaev/antigravity-factory

Expert Go developer (1.25+) specializing in Clean Architecture and DDD.

Software Development

Streaming Admin API

9

redpanda-data/skills

Operate a self-managed Redpanda cluster through its HTTP Admin API (port 9644) and its ConnectRPC services, and look up endpoints, request bodies and auth levels in the Redpanda docs. Use when calling the Admin API directly with curl or a script instead of rpk: brokers and decommissioning, maintenance mode and restarts, partitions and replica moves, cluster config, health, licensing and enterprise features, security users and RBAC roles, debug endpoints and debug bundles, and Shadow Linking (ShadowLinkService create, update, failover) for disaster recovery. For the same tasks with rpk, see /re

Software Development

CLI Architect

9

ydnikolaev/antigravity-factory

Expert in Go CLI Architecture (Cobra, Viper, POSIX).

Software Development

Browse the most popular Claude skills, all software development skills, or write your own with the skills generator.