---
name: irregularnet-host
description: Help an operator assess equipment, plan and configure an approved IrregularNet community Wi-Fi location, or plan a separate certificate Wi-Fi realm. Use for AI-assisted host onboarding, router and access-point setup, isolation, acceptance tests and recovery.
---

# Host IrregularNet with an AI assistant

Help the operator produce a working, approved certificate Wi-Fi location on their
own equipment. This file is self-contained: it can be attached or pasted into
Codex, Claude or another assistant, or placed in that assistant's skill folder.
It grants no access to equipment or community infrastructure.

Start with the operator's existing equipment, stated goal and permissions. Make
useful progress on assessment and configuration previews while missing approvals
or platform inputs are obtained. Do not describe a plan or a healthy router as a
working IrregularNet deployment.

## Establish the current operating model

Read the maintained sources before selecting commands or claiming availability:

- Hosting guide: https://wifi.irregulars.io/setup/hosting/
- Public hosting guide: https://irregularpedia.org/community/irregularnet-hosting/
- Roadmap: https://git.juntogroups.org/irregulars/irregularchat-monorepo/-/blob/main/selfhost/irregularnet/ROADMAP.md
- Independent realm: https://irregularpedia.org/server-guides/irregularnet-realm/
- Technical architecture: https://irregularpedia.org/server-guides/certificate-wifi-organization/

Some sources require sign-in. Ask the operator to complete authentication; do
not reuse an old OAuth callback or ask them to paste cookies, codes or tokens.
If current sources cannot be read, identify that limitation and continue with
an equipment assessment; leave activation readiness unverified.

Choose the track from the operator's goal:

- **Shared IrregularNet location:** eligible members use their own device
  certificates at approved locations. The community platform owns identity,
  issuance, certificate trust and global revocation. A host supplies equipment,
  an isolated network, approved authentication transport and onsite support.
- **Independent realm:** the organization owns its membership, trust, issuer,
  RADIUS policy and recovery. Read the realm plan and verify installer readiness
  before generating installation commands. The current repository contains Muse
  assumptions; changing an SSID or copying its bootstrap is not a new realm.

At this skill's publication, expansion and independent-realm installation were
roadmap work rather than self-service activation. Recheck current evidence on
each use. A guide, account, certificate, shared SSID or successful ping does not
approve a new location or establish federation.

## Collect the inputs that change the setup

Infer what is already available from the operator's files and authorized
read-only access. Ask for missing essentials together:

- Location/realm track, responsible operator and existing platform approval.
- Exact AP/router/controller model, hardware revision, firmware and management
  method; whether configuration access and live changes are authorized.
- Existing staff/guest networks, WAN, VLAN/switch layout, DHCP/DNS, IPv4/IPv6,
  management access and a recovery path that survives a Wi-Fi mistake.
- Intended coverage/client capacity, onsite tester and suitable change window.
- The platform-issued site configuration, if available: exact SSID, realm,
  authentication endpoints/ports, expected server identities, private transport
  and actual registered NAS source. Reference secret locations privately.

Do not request certificate profiles, client private keys, CA private keys or
RADIUS secrets in chat. Inspect secret-bearing configuration privately and
summarize names, redacted diffs or equality checks. Do not print raw exports or
embed secrets in generated commands, screenshots, tickets or worksheets.

## Assess and prepare a concrete configuration

Use official documentation for the exact installed hardware and firmware.
Research unfamiliar or changed commands instead of extrapolating from another
vendor or an old release. A WPA2/WPA3 **Personal** password-only mode is not
external-RADIUS Enterprise support. The AP forwards 802.1X/EAP-TLS to the
approved authentication service; it does not need a member's private key.

Prepare the smallest equipment-specific change that provides:

- The platform-approved Enterprise SSID and external RADIUS settings. Use the
  issued values; do not invent an endpoint, port, realm, identity or shared
  secret, or reuse Muse's private configuration.
- A separate member VLAN or equivalent segment, with explicit DHCP/DNS and
  internet policy. Check IPv4 and IPv6 paths to staff, management, other private
  networks and other clients against the agreed isolation policy.
- An approved private authentication path and distinct, registered NAS transport
  source. A public NAT address shared by sites, an outer username, client MAC or
  a user-supplied NAS label is not trusted site/member identity. Do not expose
  plaintext RADIUS on the public internet or invent a tunnel into the platform.
- Protected management access, the existing staff/guest service, and a tested
  recovery path. Back up the affected configuration privately before applying.

A shared-location host needs the platform's approved per-site credentials and
loaded site binding. The host does not need community signing keys, member
private keys or a copy of the enrollment database. Do not create a site by
editing platform database rows, changing SSO groups or bypassing admission.

Provide a non-secret site plan, exact configuration preview, affected networks,
expected interruption, verification commands and equipment-specific rollback.
Keep unresolved values explicit. If approval or provisioning is unavailable,
finish these artifacts and produce a concise handoff to the responsible
platform operator. Do not fabricate an activation API or run a pilot-only
bootstrap against the community service.

## Apply within the authorized scope

Reuse permissions already granted for the stated work. When live changes are
not yet authorized, obtain approval for the concrete preview after it is ready.
Local router permission does not confer community-platform administration.

Apply through the equipment's supported controller or configuration mechanism.
For remote work, establish and verify an appropriate recovery mechanism before
changing the path carrying the management session; otherwise stage the change
for an onsite operator. Preserve a record of the specific operation and result.

After a timeout or lost connection, inspect that operation's current state via
an authorized read-only path. Report succeeded, failed or unknown; do not replay
an uncertain configuration write. Roll back the scoped network change when
appropriate, rather than restoring community certificate/enrollment state.

## Prove the location works

Coordinate these checks with the onsite tester and platform operator:

1. An eligible member enrolls each test device separately through the approved
   portal. One member can have several device credentials; do not distribute one
   person's certificate/private key as a shared venue login.
2. The client validates the approved server CA and exact server name, completes
   real EAP-TLS authentication and joins the intended member network. Never turn
   off server verification to make a connection succeed.
3. Check DHCP/DNS, useful internet/application traffic and the agreed IPv4/IPv6
   isolation, including blocked staff and management access.
4. Test reconnect, reauthentication and relevant sleep/wake or roaming behavior.
   Coordinate denied/revoked test credentials with their authorized owner and
   platform administrator; do not revoke a real member's existing certificates
   as an incidental setup test.
5. Measure existing-session behavior after revocation or restriction. Neither
   implies instant disconnection. A site operator's restriction is local to
   their assigned location and requires a comment; global certificate revocation
   belongs to platform administrators. Test cross-location behavior only where
   approved locations and that workflow actually exist.
6. Exercise the agreed authentication-outage and network rollback paths. For an
   independent realm, prove recovery preserves revocations newer than a backup.

Keep source/image and hardware/firmware versions, redacted configuration
fingerprints, test devices/OS, expected versus observed results, unresolved
limitations and the responsible operator. Mark each unperformed test as pending.
Configuration validation, CI or a RADIUS acceptance alone is not onsite client
acceptance. Activation follows the actual platform and host approval process.

## Finish with an operator handoff

State the real outcome: assessed, prepared, configured, onsite-tested or
activated. Supply the useful artifacts already produced, where private backup
and recovery instructions are stored, current limitations and the next owner
for any remaining step. Do not publish the location or contact details, submit
an application or send messages on the operator's behalf without authorization.

A useful starting request is:

> Read the attached IrregularNet hosting skill and help me set up a community
> location. My AP/router model and firmware are [provide them]. I want [a shared
> IrregularNet location or an independent realm]. Start with the current roadmap
> and a read-only assessment, then prepare a concrete configuration preview.
