Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Network topology discovery via CDP/LLDP neighbors, ARP tables, routing peers, and interface mapping to build complete network maps. Use when mapping the network, building a diagram, discovering what is connected to what, or documenting device neighbors and links.
.claude/skills/automateyournetwork-pyats-topology/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 163% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 174% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 167% | 0% |
bashPYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "${PYATS_PYTHON:-python3} -u $PYATS_MCP_SCRIPT" pyats_run_show_command '{"device_name":"R1","command":"show cdp neighbors detail"}'
Extract per neighbor:
Build adjacency table:
Local Device | Local Interface | Remote Device | Remote Interface | Remote Platform
R1 | Gi0/0/0 | SW1 | Gi1/0/1 | WS-C3850-24T
R1 | Gi0/0/1 | R2 | Gi0/0/0 | ISR4431bashPYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "${PYATS_PYTHON:-python3} -u $PYATS_MCP_SCRIPT" pyats_run_show_command '{"device_name":"R1","command":"show lldp neighbors detail"}'
LLDP is IEEE 802.1AB — works with non-Cisco devices (Arista, Juniper, Linux hosts, IP phones, APs). Same adjacency table format as CDP but may include additional TLVs.
bashPYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "${PYATS_PYTHON:-python3} -u $PYATS_MCP_SCRIPT" pyats_run_show_command '{"device_name":"R1","command":"show arp"}'
Analysis:
OSPF neighbors = L3 adjacent routers:
bashPYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "${PYATS_PYTHON:-python3} -u $PYATS_MCP_SCRIPT" pyats_run_show_command '{"device_name":"R1","command":"show ip ospf neighbor"}'
BGP peers = logical connections (may be multi-hop):
bashPYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "${PYATS_PYTHON:-python3} -u $PYATS_MCP_SCRIPT" pyats_run_show_command '{"device_name":"R1","command":"show ip bgp summary"}'
EIGRP neighbors:
bashPYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "${PYATS_PYTHON:-python3} -u $PYATS_MCP_SCRIPT" pyats_run_show_command '{"device_name":"R1","command":"show ip eigrp neighbors"}'
bashPYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "${PYATS_PYTHON:-python3} -u $PYATS_MCP_SCRIPT" pyats_run_show_command '{"device_name":"R1","command":"show ip interface brief"}'
Build subnet map:
Interface | IP Address | Subnet | Connected Subnet
Gi0/0/0 | 10.1.1.1/30 | 10.1.1.0/30 | R1 <-> SW1 transit
Gi0/0/1 | 10.1.2.1/30 | 10.1.2.0/30 | R1 <-> R2 transit
Loopback0 | 1.1.1.1/32 | 1.1.1.1/32 | Router IDbashPYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "${PYATS_PYTHON:-python3} -u $PYATS_MCP_SCRIPT" pyats_run_show_command '{"device_name":"R1","command":"show vrf"}'
For each VRF, identify:
bashPYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "${PYATS_PYTHON:-python3} -u $PYATS_MCP_SCRIPT" pyats_run_show_command '{"device_name":"R1","command":"show standby brief"}'
Map virtual IPs, active/standby roles, group numbers, and tracking objects.
Combine all discovery data into a unified model:
Topology: NetClaw Discovery - YYYY-MM-DD
Devices:
R1 (C8000V, IOS-XE 17.x.x)
Loopback0: 1.1.1.1/32 (Router ID)
Gi1: 10.1.1.1/30 → R2:Gi1 (OSPF Area 0, cost 1)
Gi2: 10.1.2.1/24 → SW1:Gi0/1 (Access VLAN 10)
R2 (ISR4431, IOS-XE 17.x.x) [discovered via CDP]
Gi1: 10.1.1.2/30 → R1:Gi1
Gi2: 10.2.1.1/24 → SW2:Gi0/1
Subnets:
10.1.1.0/30 - R1-R2 transit (OSPF Area 0)
10.1.2.0/24 - R1 LAN segment (VLAN 10)
10.2.1.0/24 - R2 LAN segment (VLAN 20)
Routing Adjacencies:
R1 <-> R2: OSPF (Area 0, FULL)
R1 <-> ISP: BGP (AS 65001 <-> AS 65000, Established)
FHRP:
VLAN 10: HSRP Group 10, VIP 10.1.2.254, Active=R1, Standby=R3After discovery, use this data to generate:
When NetBox is available ($NETBOX_MCP_SCRIPT is set), reconcile discovered topology against the source of truth:
bashpython3 $MCP_CALL "python3 -u $NETBOX_MCP_SCRIPT" netbox_get_objects '{"object_type":"dcim.cables","filters":{},"limit":200}'
bashpython3 $MCP_CALL "python3 -u $NETBOX_MCP_SCRIPT" netbox_get_objects '{"object_type":"dcim.devices","filters":{},"brief":true}'
bashpython3 $MCP_CALL "python3 -u $NETBOX_MCP_SCRIPT" netbox_get_objects '{"object_type":"dcim.interfaces","filters":{"device":"R1"}}'
Compare CDP/LLDP discovered neighbors against NetBox cables:
| Category | Meaning | Action | |---|---|---| | DOCUMENTED | Link exists in both discovery and NetBox | No action | | UNDOCUMENTED | Link found by CDP/LLDP but not in NetBox | Open ServiceNow incident to update NetBox | | MISSING | Cable in NetBox but not seen by CDP/LLDP | Investigate — may be physical disconnect | | MISMATCH | Endpoints differ between discovery and NetBox | Investigate — possible re-patching |
Generate a Draw.io topology diagram with links color-coded by reconciliation status:
Run CDP/LLDP/ARP/routing peer collection across ALL devices simultaneously using multiple exec commands. Merge results to build the complete topology graph.
Record the topology discovery in GAIT:
bashpython3 $MCP_CALL "python3 -u $GAIT_MCP_SCRIPT" gait_record_turn '{"input":{"role":"assistant","content":"Topology discovery completed: 5 devices, 12 links. NetBox reconciliation: 10 documented, 1 undocumented, 1 missing.","artifacts":[]}}'
Other measured skills in the registry, with their headline benchmark lift.