# Xr Lab Assistant > Building or modifying IOS-XR/XRd labs or network topologies, creating router configurations, starting or stopping testbeds, running show commands, exploring protocol behavior, or troubleshooting boot and config issues. Use when this capability is needed. - Skill: `tomevault-io/xr-lab-assistant` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/xr-lab-assistant` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/xr-lab-assistant/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/xr-lab-assistant --- # XR Lab Assistant ## Overview Act as an IOS-XR networking virtual lab assistant for this repo. Create or update lab files, launch/debug the lab using the lab backend tooling, and run show or verification commands against running labs. You are working with IOS-XR platforms. IOS-XR syntax differs from IOS XE and other Cisco operating systems; wrong syntax produces misleading errors—check the command reference when uncertain. ## Interaction Policy All user interaction requirements are defined in [references/interaction.md](references/interaction.md) and are mandatory in all duties. ## Related Skills This skill depends on lab backend skills for tooling. ### Lab Backend Lab backend skills provide commands for launching, stopping, and interacting with labs. They may have their own compute infrastructure dependencies (e.g., Docker environment setup) — follow the lab backend skill's instructions for any prerequisites. **Selection:** - If only one lab backend skill is available, use it. - If multiple lab backend skills are available: - For an existing lab, infer the backend from its topology files - For a new lab or ambiguous context, ask which backend to use with a brief recommendation - Check the selected skill for a Caveats section or `references/caveats.md` ## Tooling Operating Principles 1. **Batch lab interactions** to reduce tool calls: * For example: * When the same output is needed from multiple routers, loop over routers in a single request * When multiple show commands are needed from one router, serialize them in a single request * If a sleep is required before interaction, combine the wait with the next interaction step * Combine these approaches whenever safe while keeping output readable ## Lab Structure A lab is a directory containing: - tooling-specific topology definition file(s) - `.cfg` — one config file per router - `README.md` — topology docs and smoke tests Labs can live anywhere — identified by a path relative to the workspace root or an absolute path. ## Duty Model Operate in exactly one duty at a time. Determine the appropriate starting duty based on the user's request, and switch as needed: - Designing - Launching - Interacting ### Context Refresh When entering a duty (whether at the start of a session or via a transition), **read** the following before proceeding: - [references/interaction.md](references/interaction.md) (always — interaction rules apply to all duties) - Designing: also read [references/designing.md](references/designing.md) - Launching: also read [references/launching.md](references/launching.md) This ensures detailed procedural and interaction requirements are in active context, especially in long sessions where earlier instructions may have faded. ### Designing Duty Use when defining a new lab or updating an existing lab. See [references/designing.md](references/designing.md) for file formats, config standards, and README requirements. ### Launching Duty Use when starting, stopping, or testing a lab. See [references/launching.md](references/launching.md) for rules, post-launch testing phases, and failure handling. ### Interacting Duty This duty is open-ended - run commands in the lab as needed to achieve the task. --- > Source: [ios-xr/xrd-tools](https://github.com/ios-xr/xrd-tools) — distributed by [TomeVault](https://tomevault.io).