Claude-Pipeline: Skill for Claude Code

Create a new database migration SQL file. Follow these steps exactly:

Step 1: Find the next migration ID

Read the current migration files in src/infrastructure/database/migrations/ and find the highest numbered migration file. The new migration ID is that number + 1.

Known duplicate IDs: Migrations 116 and 133 have duplicates — be aware of this but it only matters if creating migrations in that range (you won't be).

Step 2: Create the migration file

Create: src/infrastructure/database/migrations/<ID>_$ARGUMENTS.sql

Example: If the highest migration is 226_add_user_quota.sql and the user wants "add_status_to_contacts", create 227_add_status_to_contacts.sql.

Step 3: Write safe, idempotent SQL

All migrations 206+ MUST be safe and idempotent. Use these patterns:

-- For new tables:
CREATE TABLE IF NOT EXISTS table_name (
  id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
  user_id UUID NOT NULL REFERENCES users(id) ON DELETE CASCADE,
  created_at TIMESTAMPTZ DEFAULT NOW(),
  updated_at TIMESTAMPTZ DEFAULT NOW()
);

-- For new columns:
ALTER TABLE table_name ADD COLUMN IF NOT EXISTS column_name VARCHAR(100);

-- For indexes:
CREATE INDEX IF NOT EXISTS idx_table_column ON table_name(column_name);

-- For constraints:
DO $$ BEGIN
  ALTER TABLE table_name ADD CONSTRAINT constraint_name UNIQUE (col1, col2);
EXCEPTION WHEN duplicate_object THEN NULL;
END $$;

Step 4: Register in migrate.js

Open scripts/migrate.js and add the new migration filename to the SAFE_MIGRATIONS array at the end.

Rules

• NEVER create destructive migrations (DROP TABLE, DROP COLUMN)
• Always use IF NOT EXISTS / IF EXISTS for idempotency
• Always include user_id references where applicable (multi-tenant)
• Use TIMESTAMPTZ for timestamps, not TIMESTAMP
• Use gen_random_uuid() for UUID defaults
• Use ON DELETE CASCADE for foreign keys to users(id)