Menu
Use when implementing search with Postgres: tsvector, GIN indexes, search RPC, or replacing client-side filter of large lists. Not for external Algolia/Typesense setup unless migrating away from DB search.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
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 | postgres-full-text-search |
| description | Use when implementing search with Postgres: tsvector, GIN indexes, search RPC, or replacing client-side filter of large lists. Not for external Algolia/Typesense setup unless migrating away from DB search. |
alter table public.posts add column search_vector tsvector
generated always as (
setweight(to_tsvector('english', coalesce(title, '')), 'A') ||
setweight(to_tsvector('english', coalesce(body, '')), 'B')
) stored;
create index posts_search_idx on public.posts using gin (search_vector);
Use simple config for mixed language products; consider pg_trgm for typo-tolerant name search.
create or replace function public.search_posts(
p_query text,
p_limit int default 20
) returns setof public.posts
language sql stable as $$
select *
from public.posts
where search_vector @@ websearch_to_tsquery('english', p_query)
and deleted_at is null
order by ts_rank(search_vector, websearch_to_tsquery('english', p_query)) desc
limit p_limit;
$$;
Grant execute to authenticated; RLS still applies on underlying table.
const { data } = await supabase.rpc("search_posts", { p_query: q });
Debounce input 300ms; minimum query length 2.
create extension if not exists pg_trgm;
create index projects_name_trgm_idx on public.projects using gin (name gin_trgm_ops);
-- where name ilike '%' || p_query || '%' — use sparingly at scale
ilike '%term%' on large tables without index.soft-delete-and-archiving)