Determine next number -- use Glob to scan the migrations directory for existing migration files and find the highest number, then increment by 1 (zero-pad to 3 digits)
Select template -- based on the <name>, choose the appropriate SQL template:
- Names starting with create_ -> CREATE TABLE template
- Names starting with add_ -> ALTER TABLE ADD COLUMN template
- Names starting with drop_ -> DROP with safety checks
- Names containing index -> CREATE INDEX template
- Other -> generic migration template with placeholder comments
Generate up migration -- write NNN_<name>.up.sql with the appropriate SQL using IF NOT EXISTS for idempotency
Generate down migration -- write NNN_<name>.down.sql with the reverse operation using IF EXISTS
Search past patterns -- call mcp__claude-flow__agentdb_pattern-search (ReasoningBank-routed; don't pass a namespace argument — pattern-* tools ignore it).
Store metadata -- call mcp__claude-flow__memory_store --namespace migrations to record the migration with number, name, status (pending), and file paths. The memory_* tool family routes by namespace; agentdb_hierarchical-* does NOT (it routes by tier working|episodic|semantic), so use memory_* here. See ruflo-agentdb ADR-0001 §"Namespace convention".
Report -- display: migration number, file paths created, template used, any similar past migrations found

CLI alternative

npx @claude-flow/cli@latest memory store --namespace migrations --key "migration-NNN_NAME" --value '{"number": NNN, "name": "NAME", "status": "pending"}'

migrate-create

Migrate Create

When to use

Steps