docker-agent/examples/k8s_debugger.yaml at main · docker/docker-agent · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
agents:
  root:
    model: openai/gpt-5
    description: An AI Assistant for Debugging Kubernetes Clusters
    instruction: |
      You are an AI assistant for debugging Kubernetes clusters. Your role is to identify and resolve issues by using diagnostic tools and analyzing cluster data.

      Instructions:
        1. Ask if the user has observed specific issues or symptoms in the cluster.
        2. Use available tools to gather information about resources or real-time activity.
        3. Analyze the collected data to detect potential problems.
        4. Highlight the potential issues and suggest actionable recommendations to resolve them.

      Important:
        - If Inspektor Gadget is not deployed, use the tool at hand to deploy it. Ask user to restart the agent session after deployment.
        - If possible keep the volume of data collected to a minimum by using specific queries and filters.
        - Always use the tools provided to gather data before making any conclusions.

    toolsets:
      - type: mcp
        ref: docker:inspektor-gadget
        ## See full if of tools: https://hub.docker.com/mcp/server/inspektor-gadget/overview
        config:
          # TODO: Replace with the path to your kubeconfig file (e.g /home/user/.kube/config)
          kubeconfig: YOUR_KUBECONFIG_PATH
          # List of gadgets to enable (see https://artifacthub.io/packages/search?kind=22&verified_publisher=true&official=true&cncf=true&sort=relevance&page=1 for a full list)
          gadget-images: trace_dns,trace_tcp,trace_tcpretrans
      - type: mcp
        ref: docker:kubernetes
        config:
          # TODO: Replace with the path to your kubeconfig file (e.g /home/user/.kube/config)
          # When running with Docker Engine (not Docker Desktop), the kubeconfig file on the host must have owner with uid 1001
          config_path: YOUR_KUBECONFIG_PATH
        tools:
          ## See the full list of kubectl commands: https://hub.docker.com/mcp/server/kubernetes/tools
          - ping
          - kubectl_get
          - kubectl_describe
          - kubectl_logs
      - type: memory
        path: "./k8s_debugger_memory.db"