delete-initial-resource

Overview

Removes one or more of the five initial Atomic CRM resources (contacts, companies, deals, tags, tasks) and every reference to them. Irreversible confirm the target(s) with the user first; rely on git to recover. Exit criterion: make typecheck && make lint is clean and no live reference to the deleted resource(s) survives in src/ or supabase/.

When to Use

• The user asks to delete, remove, or strip out one or several built-in CRM resources.
• Not for custom entities added during setup only the five built-ins above.

For the backend migration mechanics, see Skill({skill: "backend-dev"}) and Skill({skill: "writing-migrations"}).

Steps

1. Confirm the target(s), then read each one's file before editing:
   • contacts → contacts.md — spine; confirm cascade scope first
   • companies → companies.md — spine/link; confirm cascade scope first
   • deals → deals.md
   • tags → tags.md
   • tasks → tasks.md
2. Run the script (below) — deletes each src/components/atomic-crm/<resource>/ folder and prints the dependent files to clean.
3. Clean each dependent file using the "Shapes" patterns below + the per-resource file.
4. Verify (below).

Deleting several at once: clean from the most-coupled resource outward; clean a shared file once for both; and reconcile interacting union guidance — deleting both contacts and deals removes the notes/activity-log reference entirely rather than each narrowing it, so don't half-apply each file's "narrow the union" note.

The script

node .claude/skills/delete-initial-resource/delete-initial-resource.ts <resource> [<resource> ...]

Validates the names, deletes each resource's folder, and prints the merged, de-duplicated dependent-file list (files inside a just-deleted folder are dropped). It does not edit those files — that's your job. Each entry in the script's dependentFiles map has a // comment saying what to remove; read it alongside the printed list, and update the map if the codebase grows new references.

Shapes a resource takes

A resource is rarely just a folder. The later shapes are invisible to a \b<resource>\b grep:

• Standalone — imports, JSX, routes, menu/nav, dashboard widgets, the <Resource> entry in root/CRM.tsx. Wrapper components outside the folder that only render the resource: delete outright, then clean importers.
• Field/column on another record — fans out far wider: the types.ts type, providers/commons/mergeContacts.ts, CSV/JSON import, i18n, stories/tests, StoryWrapper builders, and the sample CSVs (header and every row). For a link column, decide per-link: orphaned shell (typed, defaulted) or drop — remove the UI either way.
• Aggregated nb_<resource> (from the _summary views) lands on a record type, read across shows/lists/filters/generators. Invisible to a \b<resource>\b grep (the _ is a word char) — grep nb_<resource> and the denormalized company_name separately.
• Shared subsystems — notes (reference: "contacts" | "deals") and the activity log ("company" | "contact" | "deal" | "all" union). Narrow the union, delete the per-type render components (ActivityLog*Created.tsx) + dead branches, prune the providers/commons/activity.ts fetcher + consts.ts/types.ts members. Surviving === "<other>" checks stay valid — only the deleted member's branches go.
• Config props (taskTypes, dealStages, companySectors) live in root/defaultConfiguration.ts, ConfigurationContext.tsx, CRM.tsx (default + jsdoc + store seed), App.tsx, and SettingsPage.tsx (the <Card> and the section-list entry). Not always named after the resource — grep the settings <Card> for other keys (Deals owned currency) and flag borderline-general ones rather than dropping them silently.
• Custom dataProvider methods live in both providers; CrmDataProvider derives from the supabase one, so removing them there drops them from the type. Also clean the lifecycle-callback blocks in both providers, the sales beforeDelete reassignment, the supabase <resource> -> <resource>_summary getList routing, and the FakeRest generators (which drive record creation).
• JSON importer (misc/useImportFromJson.ts) treats contacts/companies/notes/tasks as top-level types (import<X> fn, $.<x>.* path, a TYPES arm + switch, guards, stats/failedImports/idsMaps keys). Mirror in ImportPage.tsx + import-sample.json; the CSV importer (useContactImport.tsx) has parallel mappings.
• Check useCallback/useMemo deps — a stale getTags/getCompanies there is invisible until tsc.

Backend (the script only touches frontend)

See the backend-dev skill for the migration workflow. In supabase/schemas/:

• Drop/adjust the table + view, generate a migration (npx supabase db diff --local -f remove_<resource>); update contacts_summary/companies_summary if they aggregated it.
• activity_log is a UNION ALL with the same column set per branch — remove the resource's columns from every branch (keep them aligned), delete its dedicated branch(es), and drop the matching snake→camel rename in the supabase getList("activity_log").
• 06_grants.sql grants the table and its <resource>_id_seq (drop both). 05_policies.sql has an RLS-enable line and create policy statements (drop both).
• A resource used inside a SQL function (merge_contacts): drop those lines + the unused local var. Editing 02_functions.sql regenerates the whole function (expected) — renumber leftover step comments.
• Orphan check: deleting triggers can orphan helpers — but verify shared callers first (get_domain_favicon survives a companies delete because contact avatars also call it).
• Edge functions are separate and easy to miss: merge_contacts/, _shared/db.ts (Kysely types), postmark/ (inbound email), and mcp/ (an AI subsystem whose depth varies per resource). When a surviving function loses a capability, flag it to the user.

Rationalizations

Red Flags

• Editing files inside supabase/migrations/ (never — they are append-only history).
• A one-sided i18n key removal (English without French, or vice versa).
• Relying on a single \b<resource>\b grep and skipping nb_<resource> / company_name.
• Deleting a shared SQL helper or edge-function capability without checking surviving callers.
• Deleting several resources but half-applying each file's "narrow the union" note.
• Skipping the db reset check when Supabase is running.

Verification

make typecheck && make lint

Resolve whatever tsc surfaces (the dependent-file list is a guide, not a guarantee). Then grep -rniE "\b<resource>\b" over src/+supabase/, plus separate greps for nb_<resource>, company_name, and config-prop names (<resource>Types, companySectors, currency). Watch for benign/substring false positives (each resource file lists its own); never edit supabase/migrations/.

• make typecheck && make lint is clean.
• \b<resource>\b, nb_<resource>, company_name, and config-prop greps return only benign matches.
• i18n keys removed from both catalogs, or neither.
• Backend: table/view/policies/grants/sequences and any function references handled in supabase/schemas/ (not migrations/).
• If Supabase is running, npx supabase db reset --local replays cleanly.
• Any lost edge-function capability flagged to the user.

i18n is all-or-nothing: frenchCrmMessages.ts is type-checked against the type derived from englishCrmMessages.ts, so remove a key from both catalogs or neither (a one-sided removal is a tsc error). Dead resources.<x>.* keys are harmless — flag rather than force. frenchCrmMessages.ts uses literal … escapes; if an exact-string Edit fails on a block spanning one, fall back to sed -i '<from>,<to>d'.

If Supabase is running, npx supabase db reset --local replays the chain + seed.sql and catches a broken view/function/grant db diff misses (benign index "…_pkey" does not exist NOTICEs expected). On Node 22 (see .nvmrc), make test (vitest) is also available.