Configure with Claude skills

Helix ships a set of Claude Code skills with the core image to help you configure your deployment. Open your Helix directory in Claude Code and the skills are available as /-commands that walk you through common tasks — grounded in the real Helix files and fields, not generic advice.

How they arrive

The skills are core-owned and travel with the helixCore image, so every deployment gets the current set:

1. They are baked into the image at build time.
2. helixctl pull (or install) materialises them into .claude/skills/ next to your node’s launcher — the directory where you manage your domain files. (Best-effort: if the image isn’t pulled yet, they install on the image is where they come from, so fetching one is what installs them.)

There is nothing to install by hand and no client-repo change — they update whenever you move to a newer core.

Using a skill

Run Claude Code from your Helix deploy directory (the one holding helix-config/), then invoke a skill by name, for example:

/customer-icons  add a yellow van icon for the RSS resource type

Claude follows the skill’s steps — which files to edit, the naming conventions, the domain field to set — and makes the change in context. Type / to see the available skills.

Available skills

• customer-icons — add or brand a map/marker icon: where the PNG lives in your client repo, the name convention, and how to point a resource or stop type at it via helix.icon (including colour-branding a silhouette per livery).
• customer-icon-artwork — draw the icon itself: SVG masters for vehicle markers and incident glyphs, framed so the map tile’s status bar and badges stay clear, with build/check scripts.
• domain-stop-styling — style stops from the domain: per-state colours and labels, clientStates vs display values, time-band colour arrays (and the "null" leave-alone band), and the off-duty roster shift periods (std/dueOn/late/noShow) with their thresholds and boundary actions — plus the gotchas that catch every new domain (JSONC stream format, fail-open expression names, band ordering).

More configuration skills are added over time and arrive automatically with each core release.

Adding your own (client teams)

If you maintain a client extension (see Develop a client extension), you can add project-specific skills in your own repo under .claude/skills/<name>/SKILL.md — a folder with a SKILL.md whose frontmatter gives a name and a one-line description (what it’s for), followed by the steps. Client-local skills sit alongside the core-shipped ones. To contribute a skill to the shared core set so every client gets it, raise it with your Honeywick contact.