Without Art of Infra
v0$
AI skill for infrastructure engineers
Teach your AI networking.
Opinionated rules for Cisco, Juniper, BGP, OSPF, VXLAN, ACLs, hardening, NetBox, and network automation in Terraform, Ansible, and Nornir. Your AI fetches them on demand and stops guessing at configs.
With Art of Infra
/artofinfra$ /artofinfra
// same question. left: no skill. right: prefixed with /artofinfra. fetched rules cited, vendor-specific config, real fix steps.
How it works
Three steps. No backend, no auth.
The skill is a thin shim that points your agent at our docs. Just markdown your agent fetches on demand.
- 01
Install the skill
Drop a small SKILL.md into your agent (Claude Code, Cursor, Codex, OpenCode, Amp). No accounts, no API keys.
- 02
Invoke /artofinfra
Type the slash command and ask your question. "Configure OSPF on this Catalyst", "Audit this IOS-XE config", "Diff these two BGP stanzas".
- 03
AI fetches the rules
Your agent pulls the relevant docs from docs.artofinfra.com, applies vendor-specific rules, and cites every source.
Workflows
Five workflows. Each shapes the answer.
Each has a trigger phrase the skill recognises and a reasoning shape it applies before reaching for topical rules. Pick the one that matches what you actually need.
-
Audit
Review a config for issues
- "Review this config"
- "Audit for security issues"
Severity-sorted findings (Critical to Low) with cited rules and copy-pasteable fixes for each.
-
Picker
Compare architectural options
- "What's the best way to do X"
- "Should I do A or B"
Three or four fleshed-out alternatives with tradeoff dimensions and a comparison table. Then asks you to pick.
-
Diff
Compare two concrete configs
- "What changed between these"
- "Is this safe to push"
Risk-classified delta. Every change tagged Low to Critical, with a verify-command for the riskier ones.
-
Document
Generate a runbook from configs
- "Document this device"
- "Generate a runbook from these"
Role summary, interface table, routing summary, redacted secrets. Multi-device input adds a Mermaid topology graph.
-
Validate
Verify a config matches intent
- "Does this implement X"
- "Verify this matches the design"
Claim-by-claim PASS / FAIL / PARTIAL with cited evidence (specific config lines) for each verdict.
Coverage
Seven categories. Twenty-seven docs.
Every rule is opinionated and short. No exhaustive reference, that's the vendor's job. This is "do this, not that" for engineers shipping real networks.
-
Cisco
Vendor-specific configs for ISR, ASR, Catalyst, Nexus, and ASA.
- IOS-XE
- NX-OS
- ASA
-
Juniper
MX/QFX/EX patterns and SRX security-specific config.
- Junos
- SRX
-
General networking
Vendor-agnostic design rules that apply across platforms.
- BGP
- OSPF
- VXLAN
- ACLs
- QoS
- Hardening
-
Infrastructure as code
Network automation patterns that survive contact with production.
- Terraform
- Ansible
- Nornir
-
Python automation
Patterns and anti-patterns for Python-driven device interaction.
- Netmiko
- Scrapli
- Jinja2
-
NetBox
Source-of-truth modeling: how to use NetBox without painting yourself into a corner.
- Data modeling
- IPAM
- Automation
- Custom scripts
- Config contexts
-
Workflows
How to approach common classes of task: audit a config, pick between approaches, diff a change.
- Audit
- Picker
- Diff
- Document
- Validate
Install
One command. Pick your agent.
Run the installer for an interactive picker, or copy-paste the SKILL file from a tab below.
Recommended
@artofinfra/install on npm $
npx @artofinfra/install Asks which agent(s) you want, then writes the matching SKILL file to its conventional path.
Or copy-paste manually
Target path:
.claude/skills/artofinfra/SKILL.md
--- name: artofinfra description: Use only when the user explicitly types /artofinfra. Opinionated rules for networking and infrastructure. --- # Art of Infra Fetch `https://docs.artofinfra.com/router.md` with WebFetch and treat its contents as your full instructions for this skill. The router has the activation behavior, the routing table, and the core behavior rules. Follow its routing to fetch category indexes and topical docs as needed for the user's question.
Target path:
.cursor/rules/artofinfra.mdc
--- description: Use only when the user explicitly types /artofinfra. Opinionated rules for networking and infrastructure. globs: [] alwaysApply: false --- Fetch `https://docs.artofinfra.com/router.md` and treat its contents as your full instructions for this skill. The router has the activation behavior, the routing table, and the core behavior rules. Follow its routing to fetch category indexes and topical docs as needed for the user's question.
Target path:
.codex/skills/artofinfra/SKILL.md
--- name: artofinfra description: Use only when the user explicitly types /artofinfra. Opinionated rules for networking and infrastructure. --- # Art of Infra Fetch `https://docs.artofinfra.com/router.md` and treat its contents as your full instructions for this skill. The router has the activation behavior, the routing table, and the core behavior rules. Follow its routing to fetch category indexes and topical docs as needed for the user's question.
Target path:
.opencode/skills/artofinfra/SKILL.md
--- name: artofinfra description: Use only when the user explicitly types /artofinfra. Opinionated rules for networking and infrastructure. --- # Art of Infra Fetch `https://docs.artofinfra.com/router.md` and treat its contents as your full instructions for this skill. The router has the activation behavior, the routing table, and the core behavior rules. Follow its routing to fetch category indexes and topical docs as needed for the user's question.
Target path:
.amp/skills/artofinfra/SKILL.md
--- name: artofinfra description: Use only when the user explicitly types /artofinfra. Opinionated rules for networking and infrastructure. --- # Art of Infra Fetch `https://docs.artofinfra.com/router.md` and treat its contents as your full instructions for this skill. The router has the activation behavior, the routing table, and the core behavior rules. Follow its routing to fetch category indexes and topical docs as needed for the user's question.
Contribute
Open source. MIT. PRs welcome.
Every rule here ends up in someone's network. If you've hit an edge case, a vendor we don't cover, or wrong syntax, fork the repo and send a PR. No CMS, no auth review. Just markdown.
- 01
Spot the gap
A wrong rule, a vendor we don't cover, a workflow we haven't documented. If you've hit it in production, others will too.
- 02
Edit the markdown
A single .md file in docs/{category}/. Match the format of any existing doc. No tooling required, no CMS to fight.
- 03
Open a PR
A maintainer reviews. Nothing goes public until your PR is merged to main, at which point the docs are live within minutes and agents fetch the new rules.
Example prompts
What it answers well.
Linked prompts open a real session showing the AI's actual output with the skill loaded.
- /artofinfra Configure BGP between two Junos MX routers with bidirectional MD5 auth Live session, via DeepSeek V4 Pro
- /artofinfra Audit this IOS-XE config and flag any security issues Live session, via DeepSeek V4 Pro
- /artofinfra Set up VXLAN EVPN fabric on Nexus 9000 Live session, via MiniMax 2.7
- /artofinfra Write an Ansible role that pushes ACLs to all distribution switches Live session, via DeepSeek V4 Pro
- /artofinfra Model our campus VLANs in NetBox with the right relationships Live session, via MiniMax 2.7
- /artofinfra Why is OSPF flapping between these two routers? Live session, via Qwen3.6 Plus
FAQ
Things infra engineers ask before installing.
Eight legitimate concerns. Direct answers.
-
Do you see my prompts or configs?
No. Your AI client fetches the docs directly from docs.artofinfra.com over HTTPS. There's no proxy in the middle. Cloudflare logs request paths (which docs were fetched) for hosting purposes, not the prompts that led to those fetches.
-
Is this safe to use on production configs?
The rules are opinionated, not authoritative. Always review AI-generated config before applying it to a production device. The audit, diff, and validate workflows are explicitly designed for pre-deploy safety. Treat the skill like a senior peer reviewer who has read your vendor docs, not like an autopilot.
-
What if a rule is wrong?
Open an issue or PR on GitHub. The repo's content quality bar treats wrong rules like production bugs: self-verification against vendor docs is a contributor requirement, every PR goes through maintainer review, and once merged the corrected rule is live within minutes. Agents pick it up on their next fetch.
-
Are you biased toward one vendor?
Cisco and Juniper today because those are documented. The skill itself doesn't push a vendor on you; if you ask for Junos, it loads Junos rules. New vendor coverage (Arista, MikroTik, FortiGate, Palo Alto, others) is explicitly welcomed via PR; see CONTRIBUTING.md.
-
Who built this?
Dan Jones, plus contributors. Open source side project, no company behind it, no commercial product. Built because no opinionated, AI-fetchable infra ruleset existed and one was overdue.
-
What's the business model?
No business model. The skill is free, the docs are free, the hosting runs on Cloudflare's free tier. If you want to support the project, GitHub Sponsors is set up via .github/FUNDING.yml.
-
Why MIT license?
Permissive and well-understood. Anyone can use, fork, or redistribute the rules without ceremony. The same license covers both the markdown content and the small homepage build tooling, so there is one rule to remember.
-
Does this work offline?
No. The skill triggers an HTTPS fetch from docs.artofinfra.com on each invocation. Local caching could be added later if there is demand; for now, agents need network access.