All six of the kit's skills work in Cursor, unchanged, with no setup step. This surprises people, so
it is worth being precise about why.

Cursor loads skills from `.agents/skills/` and `.cursor/skills/` in a project, and from the matching
directories under your home directory. It also loads them from `.claude/skills/` and `.codex/skills/`
for backward compatibility. The kit's skills live in `.claude/skills/<name>/SKILL.md`, so they are
picked up by that last path. There is nothing to copy, symlink, or convert.

The format lines up too. Cursor requires `name` and `description` in a `SKILL.md` front matter, which
is exactly the two fields each of the kit's skills declares. Nothing further is needed.

## Three ways to reach one

- **Describe the task and let the agent choose.** Each skill's `description` is written to match how
  someone would actually phrase the underlying request, so "add invoices, one per project, with an
  amount and a due date" reaches `crud-with-billing` without you naming it. The patterns in
  [Prompting Patterns](/docs/claude-code/prompting-patterns) apply here as written.
- **Type `/` in Agent chat**, then search by name. Use this when your request could plausibly match
  more than one skill, or when you are repeating a skill you already ran for a similar feature.
- **Pin one for the session** with Option and Enter on a Mac, or Alt and Enter on Windows, which
  keeps the skill active for the whole session rather than a single message.

## The six

Run **`bootstrap-app`** first on a fresh clone, before building any feature. It renames the Rails
module, sets the product identity and brand, regenerates the icon and share image, and removes the
example `Project` slice and the kit-sale flow. Skip it and you build on top of, and eventually ship,
the kit's own scaffolding. See [The bootstrap-app Skill](/docs/claude-code/skill-bootstrap-app) and
[Making It Yours](/docs/getting-started/making-it-yours).

After that, in the order you are likely to need them:

- **`crud-with-billing`** scaffolds a new account-scoped resource: model and migration with
  `account_id`, a Pundit policy and scope, a controller querying through `Current.account`,
  Terminal-styled Hotwire views, routes, and model, policy and request specs. Optionally gated behind
  a subscription. See [The crud-with-billing Skill](/docs/claude-code/skill-crud-with-billing).
- **`add-mailer`** adds a transactional email, wired through the existing Resend and letter_opener
  setup. **`add-job`** adds a Solid Queue background job, optionally recurring. See
  [The add-mailer and add-job Skills](/docs/claude-code/skill-add-mailer-and-add-job).
- **`add-integration`** wires a new third-party service as an adapter and fake seam, so the app keeps
  running with zero external accounts. See
  [The add-integration Skill](/docs/claude-code/skill-add-integration).
- **`write-post`** publishes a blog post with the front matter and structure that earn search traffic
  and citations. See [The write-post Skill](/docs/claude-code/skill-write-post).

Those walkthroughs are written against Claude Code, because that is where the skills came from. The
steps they describe are the same in Cursor; only the way you invoke the skill differs.

## When a skill does not fire

The skills are plain markdown, so the fallback is always available: open the `SKILL.md` file in a tab
and tell the agent to follow it. Context that is already on screen is context it does not have to go
looking for, and this is the cheapest fix for a skill that the agent did not think to reach for.

## Next

Put it together on a real feature:
[Building a Feature in Cursor](/docs/cursor/building-a-feature-in-cursor).