メニュー
Use when planning a session in Lovable, breaking a feature into prompts, asking Lovable to refactor or debug, or writing a backlog of Lovable prompts. Not when actually executing inside Lovable — this guides how to phrase work, not the work itself.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use when reviewing or fixing accessibility: keyboard navigation, focus order, labels, color contrast, screen reader support, or ARIA attributes. Not for SEO-only or performance-only audits.
Use when adding product analytics (PostHog, GA4, Plausible), event tracking, funnels, or conversion measurement. Not for server logs only or A/B infra unless events are part of the task.
Use when standardizing API and edge function errors: HTTP status codes, error JSON shape, logging, or mapping Supabase/Stripe errors to client responses. Not for UI toast copy only or form field validation messages.
Use when recording audit trails: who changed what, admin action logs, compliance history, or append-only event tables. Not for product analytics (use analytics-events) or debug console logs only.
Use when scheduling background work: cron jobs, digest emails, cleanup tasks, pg_cron, Supabase scheduled edge functions, or job queues. Not for synchronous user-facing API in the request path.
Use when adding dark mode, theme toggle, system theme support, or fixing hardcoded colors that break in dark mode. Not for design-token migration unrelated to dark mode.
| name | lovable-prompting-playbook |
| description | Use when planning a session in Lovable, breaking a feature into prompts, asking Lovable to refactor or debug, or writing a backlog of Lovable prompts. Not when actually executing inside Lovable — this guides how to phrase work, not the work itself. |
Lovable does best with small, specific, verifiable prompts. This skill helps you (or another agent) write prompts that ship.
[Context] What part of the app, which files or routes.
[Goal] What "done" looks like in user-visible terms.
[Approach] Optional. Pick stack / pattern if it matters.
[Limits] What NOT to change.
[Verify] How you'll check it works.
Example:
On the
/dashboardroute, add a "Revenue (30d)" KPI card to the top row, usinguseRevenue30d(already exists). Show value, % delta, and a sparkline. Do not touch the other cards. I'll verify by loading the dashboard and comparing the number to Stripe.
Break a feature into 3–7 prompts, each shippable on its own:
migration-playbook).typed-api-hooks-forms).error-states-and-empty-ui).After each prompt, review the chat diff before moving on.
X unchanged."path/to/file.tsx."/refactor-safe-diff) to constrain scope.| Task | Skills to invoke |
|---|---|
| New feature with Supabase | /supabase-database-design then /typed-api-hooks-forms |
| New form | /typed-api-hooks-forms (covers Zod + RHF) |
| Stripe webhook | /stripe-payment-webhooks + /payment-webhook-idempotency |
| Before launch | /lovable-ship-checklist |
| Slow page | /performance-budget |
useFoo hook; do not create a duplicate."/refactor-safe-diff with explicit files.any, inline Supabase in components).prompts.md) so they can be re-run on a new branch.