network-state-query

name	network-state-query
description	Return a focused, structured health summary for one fabric node (interfaces and BGP), instead of a raw tool dump
state	active

Network state query

Use this skill when someone asks "is <node> healthy?" or "what is the state of <node>?" for a node in the Skills & Specs Lab fabric (spine1, spine2, leaf1, leaf2). This is the reference skill: it sets the quality bar for the skills you build in Lab 1. Study its shape before writing your own.

The point of a skill over a raw tool: a raw tool (for example the clab inspectLab tool) hands back a firehose of generic data and leaves you to find the answer. This skill asks the operational question directly and returns a scoped, labeled, structured result.

Scope

• In scope: operational state of a single named node, interfaces and eBGP.
• Out of scope: changing configuration, querying multiple nodes at once (call this once per node), anything outside the lab fabric.

Steps

1. Validate the node name is one of spine1, spine2, leaf1, leaf2. If not, stop and say so.
2. Resolve the container name: clab-skills-specs-lab-<node>.
3. Query interface state: docker exec clab-skills-specs-lab-<node> sr_cli "show interface brief"
4. Query BGP neighbor state: docker exec clab-skills-specs-lab-<node> sr_cli "show network-instance default protocols bgp neighbor"
5. Parse both. Do not paste raw CLI output back; extract the fields below.

Output

Return exactly this structure (JSON), nothing more:

{
  "node": "leaf1",
  "interfaces": [
    {"name": "ethernet-1/1", "admin": "enable", "oper": "up"},
    {"name": "ethernet-1/2", "admin": "enable", "oper": "up"},
    {"name": "system0",      "admin": "enable", "oper": "up"}
  ],
  "bgp": [
    {"peer": "10.1.1.1", "peer_as": 65100, "state": "established", "uptime": "0:12:30"},
    {"peer": "10.2.1.1", "peer_as": 65100, "state": "established"}
  ],
  "healthy": true,
  "summary": "leaf1: 3/3 interfaces up, 2/2 BGP sessions established."
}

healthy is true only when every fabric interface is up and every BGP session is established. The one-line summary is what a human reads first.

Failure modes

• Node not in the fabric: return {"error": "unknown node <name>", "valid": ["spine1","spine2","leaf1","leaf2"]}.
• Container not running: return {"node": "<node>", "error": "container not running", "hint": "./scripts/deploy.sh"}.
• BGP still converging (sessions in active/connect): report healthy: false and note in summary that the fabric may still be converging; suggest re-running after 30 seconds.