How to use the Linkerd Service Mesh Patterns agent skill

What the skill makes the agent do

Linkerd runs a small proxy alongside each application on a Kubernetes cluster. The proxies encrypt and manage traffic between services. This skill gives the agent patterns for working with that setup: installing the Linkerd CLI and control plane, enabling automatic mTLS, splitting traffic for canary releases, configuring service profiles for per-route metrics, retries and timeouts, and writing server and authorization policies that limit which services may talk to each other.

The skill is documentation and templates. It does not connect to a cluster by itself, and it does not run anything on its own.

• Installation and validation commands for the Linkerd CLI and control plane
• Namespace and deployment injection annotations
• ServiceProfile, TrafficSplit, Server, ServerAuthorization and HTTPRoute examples
• Multi-cluster setup commands
• Monitoring commands such as linkerd viz top and linkerd viz routes
• Debugging commands such as linkerd check --proxy and linkerd viz tap

How to install the linkerd-patterns skill

Download the ZIP from this page. It contains the skill's SKILL.md file. Unzip it into one of these locations, depending on how you want to use it:

• Personal use across projects: unzip into ~/.claude/skills/
• One project only: unzip into the project's .claude/skills/ folder
• Claude apps: upload the skill under Settings → Capabilities → Skills

Know what you need before you start

The skill assumes you already have a Kubernetes cluster and kubectl access to it. The agent can write the commands, but it cannot reach your cluster. The Linkerd CLI is also required. The skill's example installs it with Homebrew, and it points to the official Linkerd installation guide for other systems.

Check that the CLI version matches your target cluster before installing. The skill recommends this explicitly.

• A Kubernetes cluster and kubectl access
• The Linkerd CLI (the skill's example uses Homebrew)
• Someone who can review generated YAML before it is applied

How to trigger the skill with natural requests

You do not need special syntax. Describe the networking task in plain words and mention Linkerd or the kind of job you want done. The skill's description covers setting up Linkerd, configuring traffic policies, and zero-trust networking with minimal overhead, so requests that match those topics are the most likely to use it.

• "Set up Linkerd on my Kubernetes cluster and check that it is healthy."
• "Send 10% of traffic for my-service to a canary version using Linkerd."
• "Add retries and a 5-second timeout for GET requests on my users API in Linkerd."
• "Write a Linkerd policy so only the frontend service account can call my-service."
• "Give me the commands to debug why traffic between two pods is failing in my Linkerd mesh."

Review the generated output before applying it

The templates in the skill are generic examples. For instance, the ingress policy uses a 10.0.0.0/8 network range as a placeholder. Replace values such as service names, namespaces, CIDR ranges and weights with your own, then test in a non-production cluster.

Run linkerd check before and after changes. The skill recommends this as a standard habit. Treat the agent's output as a draft to review, not a finished change.

• Replace example names, namespaces and network ranges with your real values
• Test in a non-production cluster first
• Run linkerd check after every change

Safety notes: scripts and network

According to the skill's listing, it runs no scripts and uses no network access. Its content is instructions and YAML templates for the agent to read. Any command the agent suggests is run only in your environment and depends on your own cluster access and the agent setup you use. Review any command before running it, especially commands that modify cluster resources, and keep authorization rules as narrow as your traffic requires.