writing-skills
writing-skills
Use when creating new skills, editing existing skills, or verifying skills work before deployment
tailnet-policyAuthor, test, and deploy Tailscale-compatible huJSON policy files for Headscale tailnets — ACLs, Grants, Tags, Auto Approvers, Tailscale SSH rules. Use when configuring access control, writing policy files, or troubleshooting connectivity issues caused by ACLs.
Independent GitHub snapshot · Automated classification · Original documentation governs
Headscale uses Tailscale-compatible policy files written in huJSON (Human JSON — standard JSON with trailing commas and // comments). Policy files control:
{action, users, ports} rules{src, dst, ip, proto, via} rulestag:dev, tag:prod)ssh.users and ssh.action| Feature | ACLs (legacy) | Grants (modern) |
|---|---|---|
| Format | {action: "accept", users: [...], ports: [...]} | {src: [...], dst: [...], ip: [...], proto: "tcp"} |
| Port filtering | Embedded in ports: ["*:*"] | Separate ip field for ports |
| Protocol filtering | Not supported | proto field (tcp, udp, icmp) |
| Destination routing | Not supported | via field for relay/exit nodes |
| Status | Deprecated by Tailscale | Current recommended syntax |
Use Grants wherever possible. The migrate-acls-to-grants.py script can convert legacy ACL files automatically.
The policy file path is configured in Headscale's config.yaml:
policy:
path: /etc/headscale/policy.hujson
After modifying the policy file, reload it on the Headscale server:
# Reload via SIGHUP
kill -HUP $(pgrep headscale)
# Or use the convenience script
./skills/tailnet-policy/reload-headscale-policy.sh
{
// Grants that allow all traffic
"grants": [
{
"src": ["autogroup:member"],
"dst": ["autogroup:member"],
"ip": ["*:*"]
}
],
// Tag ownership
"tagOwners": {
"tag:dev": ["autogroup:admin"],
"tag:prod": ["autogroup:admin"]
}
}
{
"grants": [
// Only allow ICMP (ping) between all members
{
"src": ["autogroup:member"],
"dst": ["autogroup:member"],
"ip": ["*"],
"proto": "icmp"
}
],
// Specific grants added per-service
"tagOwners": {
"tag:monitor": ["autogroup:admin"]
}
}
{
"grants": [
// Dev can reach dev
{
"src": ["tag:dev"],
"dst": ["tag:dev"],
"ip": ["*:*"]
},
// Prod can reach prod
{
"src": ["tag:prod"],
"dst": ["tag:prod"],
"ip": ["*:*"]
},
// Admin access to all
{
"src": ["autogroup:admin"],
"dst": ["tag:dev", "tag:prod"],
"ip": ["*:*"]
}
],
"tagOwners": {
"tag:dev": ["autogroup:admin"],
"tag:prod": ["autogroup:admin"]
}
}
Tags are node-level identifiers set via tailscale up --advertise-tags=tag:dev. They decouple policy from user identity.
{
"tagOwners": {
"tag:ci-runner": ["autogroup:admin"],
"tag:database": ["autogroup:admin"],
"tag:webserver": ["autogroup:admin"],
"tag:monitoring": ["autogroup:admin"]
},
"grants": [
{
"src": ["tag:monitoring"],
"dst": ["tag:webserver", "tag:database"],
"ip": ["*:*"]
},
{
"src": ["tag:webserver"],
"dst": ["tag:database"],
"ip": ["tcp:5432"]
}
]
}
Grants are the modern policy primitive:
{
"grants": [
{
"src": ["tag:source", "user@example.com"],
"dst": ["tag:destination", "100.64.0.1"],
"ip": ["*:*"], // proto:port — "*:*" means all
"proto": "tcp", // optional protocol filter
"via": ["tag:exit-node"] // optional via/routing
}
]
}
Fields:
src — Source entities (tags, users, autogroups, IPs)dst — Destination entitiesip — Protocol and port filter (e.g. tcp:80, udp:53, *:*, *)proto — Protocol constraint (tcp, udp, icmp)via — Route through a specific exit node or relayAuto-approvers let specific users approve subnet routes and exit nodes without manual intervention:
{
"autoApprovers": {
"routes": {
"10.0.0.0/8": ["autogroup:admin"],
"172.16.0.0/12": ["alice@example.com"]
},
"exitNode": ["autogroup:admin"]
}
}
routes: Maps CIDR ranges to lists of users who can auto-approve those routesexitNode: Lists users who can advertise exit nodesAutogroups are dynamic groups resolved by Headscale/Tailscale at runtime:
| Autogroup | Description |
|---|---|
autogroup:member | All tailnet members |
autogroup:admin | Tailnet admins |
autogroup:tagged | All tagged nodes (any node with at least one tag) |
autogroup:internet | The public internet (used for exit node routing) |
Tailscale SSH rules are configured via the ssh section:
{
"ssh": [
{
"action": "accept", // "accept" or "check"
"src": ["autogroup:admin"],
"dst": ["tag:webserver"],
"users": ["root", "ubuntu"]
},
{
"action": "check", // "check" requires node-level SSH authorization
"src": ["autogroup:member"],
"dst": ["tag:dev"],
"users": ["*"]
}
]
}
action: "accept" (allow directly) or "check" (require node-level auth)src: Source users/groupsdst: Destination tags/usersusers: Which OS users can be SSH'd intoPolicy files include test definitions that are validated when loaded:
{
"grants": [...],
"tests": [
{
"src": "alice@example.com",
"dst": "tag:webserver",
"ip": ["tcp:443"],
"action": "accept" // expected result
},
{
"src": "bob@example.com",
"dst": "tag:database",
"ip": ["tcp:22"],
"action": "drop" // expected result
}
]
}
Validate tests with:
./skills/tailnet-policy/validate-policy.py --policy policy.hujson
devicePosture or device:managed are Tailscale-onlyipSets and ipprotocol are not supportedautogroup:admin and autogroup:member insteadtag: and contain only lowercase letters, numbers, and hyphensusers field and Grant src field are NOT interchangeable — grants use src/dst, legacy ACLs use users/portsThis skill is automatically loaded when the user's message contains any of these keywords:
Do not use this skill for deploying the Headscale server (load headscale-deploy instead) or for client connectivity issues unrelated to access control (load tailscale-client). It covers huJSON policy authoring and testing only.