Claude Code Marketplace Now Available

• Thu, Feb 19, 2026
• 5-minute read

Claude Code has become an indispensable part of my daily workflow. I use it for everything from writing code to debugging production issues. But while Claude is incredibly capable out of the box, there are areas where injecting specialized domain knowledge makes it dramatically more useful.

That’s why I built a plugin marketplace. Yesterday I released rustyrazorblade/skills, a collection of Claude Code plugins that extend Claude with expert-level knowledge in specific domains. The first plugin is something I’ve been talking about doing for a while: a Cassandra expert.

Why Build a Plugin Marketplace?

Claude is a general-purpose AI. It knows a lot about Cassandra, but it also makes up a lot of nonsense that can land you in serious trouble. I’ve seen what causes clusters to fall over during a massive traffic spike, and I’ve learned the hard way which configuration settings actually matter.

Plugins let me inject that expertise directly into Claude. Instead of Claude giving (often questionable) advice, it can apply the same systematic approaches I use when diagnosing problems or optimizing performance. The result is more useful, more targeted guidance.

I’m sharing this as a marketplace because I think others should be able to do the same thing. If you’re an expert in a domain, you can create a plugin that captures your methodology and makes it available to everyone using Claude Code.

The Cassandra Expert Plugin

The cassandra-expert plugin provides specialized guidance for Apache Cassandra systems. It’s organized into four skills, each focused on a specific type of task:

Diagnose handles systematic troubleshooting. When something goes wrong with a Cassandra cluster, there’s a methodical way to identify the root cause. I plan on updating the skill with common causes of failures to give Claude more structure in its responses, along with symptoms, and explanations as to why things happen the way they do.

Optimize covers performance tuning, configuration review, JVM optimization, and compaction strategies. Cassandra has hundreds of configuration knobs, but only a handful really matter for most workloads. This skill knows which ones to focus on.

Data Model assists with schema design, partition key selection, time-series patterns, and query optimization. Good data modeling is an absolutely critical factor in Cassandra performance, and this skill encodes the patterns that work.

Expert handles general Cassandra questions, CQL analysis, and best practices. It’s the catch-all for anything that doesn’t fit neatly into the other categories.

Problem-Solving Methodologies

The plugin doesn’t just answer questions, it applies specific problem-solving methodologies:

Double Loop Learning goes beyond fixing symptoms to identify root causes. When a cluster has high CPU usage, the surface fix might be to expand the cluster or use bigger instances. But the real fix is to identify the root cause. For example, you might be using STCS with Cassandra 5.0 and discover you’re spending a ton of CPU time on incremental repair due to anti-compaction overhead. The real solution isn’t to throw more hardware at it, it’s to migrate to UCS. I covered this in detail in my post on optimizing repair for higher node density.

USE Method, created by Brendan Gregg, provides systematic analysis of Utilization, Saturation, and Errors across system resources. Before diving into Cassandra-specific issues, this method checks whether the underlying infrastructure is the bottleneck. Is CPU saturated? Is disk I/O maxed out? Are there network errors? I’ve shared previous posts where I show how to use async-profiler and eBPF tools to determine this. These are the same tools I used with easy-cass-lab to evaluate CASSANDRA-15452, a storage engine patch I worked on with Jordan West that delivered a massive improvement in compaction throughput.

Outlier Analysis compares node behavior to detect anomalies. In a cluster, nodes should behave similarly. When one node has 10x the latency of others, that’s a signal. Outlier analysis identifies these differences and investigates why they exist.

Configuration Analysis reviews system and JVM settings against known good configurations. A classic example of this was the old JVM default of using a really small new gen, causing excess promotion, and long, frequent pauses, driving up latency.

Opinionated Recommendations

The plugin includes opinionated guidance based on real-world experience. Here’s a handful of examples:

• Keep num_tokens at 1-4. The default of 16 or 256 creates unnecessary complexity during operations.
• Use UCS compaction for Cassandra 5.0+. It’s the right choice for almost every workload.
• Disable read-ahead on SSDs. The default Linux settings assume spinning disks and hurt SSD performance.
• Keep partitions under 10MB. Larger partitions cause repair problems and uneven load distribution.
• Keep row cache disabled. It rarely helps and often hurts.

How to Install

Installation takes two commands. First, add the marketplace:

/plugin marketplace add rustyrazorblade/skills

Then install the plugin:

/plugin install cassandra-expert@rustyrazorblade-plugins

Once installed, the four skills (diagnose, optimize, data-model, expert) are available in Claude Code.

What’s Next

This is just the beginning. I plan to expand the marketplace with additional plugins covering other areas I work with regularly. I’ll also be adding more depth to the Cassandra expert plugin as I encode more of my troubleshooting knowledge.

If you’re interested in contributing, the repository is open source under the Apache 2.0 license.

Check out the repository on GitHub, try the cassandra-expert plugin, and let me know what you think.