DB Migrate

Cross-Platform AI Agent Skill This skill works with any AI agent platform that supports the skills.sh standard.

Create, validate, and run database migrations across any framework. Auto-detects the migration tool from project files.

Anti-Hallucination Guidelines

CRITICAL: Migration operations can be irreversible in production:

1. Detect before acting — Always scan for marker files before running any command
2. Never invent commands — Only use commands listed in the detection table or discovered in project docs
3. Warn on destructive operations — Always flag DROP, TRUNCATE, or irreversible schema changes
4. Dry-run first — Use --dry-run mode when supported by the framework
5. Read existing migrations — Check naming conventions before creating new ones

Workflow

Phase 1: Detect Framework

Scan for marker files to identify the migration framework:

# Check for common marker files
ls alembic.ini 2>/dev/null
ls prisma/schema.prisma 2>/dev/null
ls knexfile.* 2>/dev/null
ls flyway.conf 2>/dev/null
ls db/migrate/ 2>/dev/null
ls manage.py 2>/dev/null
ls atlas.hcl 2>/dev/null

Also check:

• package.json for knex, db-migrate, sequelize, typeorm dependencies
• pyproject.toml / requirements.txt for alembic, sqlalchemy
• Gemfile for activerecord

If multiple frameworks detected, ask user which one to use. If none detected, ask user to specify.

Refer to references/framework-commands.md for the full detection table and per-framework commands.

Phase 2: Parse Arguments

Determine operation mode from $ARGUMENTS:

• create (or no argument): Create a new migration file
• status: Show pending and applied migrations
• validate: Check rollback scripts, naming conventions, index coverage

If --name <name> provided, use it for migration name. Otherwise ask for a descriptive name. If --dry-run provided, show what would be created/run without executing.

Phase 3: Execute Operation

create — New Migration

1. Check naming conventions in existing migrations (Glob migrations/, db/migrate/) — infer pattern
2. Generate timestamp prefix: date +%Y%m%d%H%M%S
3. Run framework-native CLI to create the migration file
4. Verify file was created — read and display it to the user
5. If autogenerate is available (Alembic, Prisma): generate from schema diff

After creation, remind user to:

• Review the generated migration for correctness
• Add rollback/down script if not auto-generated
• Add indexes for any new foreign keys

status — Pending Migrations

Run the framework's status command and display:

• Applied migrations (with timestamps)
• Pending migrations (not yet applied)
• Current database schema version

validate — Migration Quality Check

Check the following:

1. Rollback exists: Every up migration has a down script (warn if missing)
2. Naming convention: All files follow the project's timestamp + description pattern
3. Foreign key indexes: Grep for REFERENCES or foreign key declarations without corresponding CREATE INDEX
4. Destructive operations: Grep for DROP TABLE, DROP COLUMN, TRUNCATE — flag each one
5. Data migrations vs schema migrations: Warn if both data and schema changes exist in one file

Phase 4: Safety Checks

Always run these before finalizing a create operation:

Destructive operation detection:

grep -iE "DROP TABLE|DROP COLUMN|TRUNCATE|DELETE FROM" <migration_file>

If found: warn user, require explicit confirmation to proceed.

Missing index check:

grep -iE "REFERENCES|foreign_key|FK_" <migration_file>

If foreign keys present, check for corresponding index creation.

Rollback verification:

• Check if the migration has both up and down sections
• For autogenerated migrations (Prisma, Alembic): verify the shadow database check passes

Phase 5: Report

Display a summary of what was created or checked:

• File path of new migration
• Migration name and timestamp
• Any warnings (destructive ops, missing rollbacks, missing indexes)
• Next steps (apply with <command>, rollback with <command>)

Argument Parsing

• create: Create a new migration (default if no subcommand given)
• status: Show migration status
• validate: Validate existing migrations
• --name <name>: Migration description for the filename
• --dry-run: Show what would happen without executing

Important Notes

• Always review auto-generated migrations before applying — they may include unintended changes
• Rollback scripts are required for production safety; warn loudly if missing
• Never run migrations directly against production without a backup
• Prisma shadow database: Prisma dev migrations use a shadow DB; ensure DATABASE_URL is not production
• Transaction wrapping: Ensure migrations run in transactions for atomicity

Examples

# Create a new migration
/db-migrate create --name add_users_table

# Check pending migrations
/db-migrate status

# Validate migration quality
/db-migrate validate

# Create without executing (preview)
/db-migrate create --name drop_legacy_column --dry-run