mcpField
S
seed

seed

github.com/hainrixzData & databases2269 installsUpdated 6d ago
Repository

About seed

seed is a read-only skill for generating deterministic sample INSERT statements from a schema, ORM model, or migration. It is intended for developers preparing reproducible fixtures for local development, testing, or CI, not production data.

Key features

  1. 01Schema-based INSERT generation
  2. 02Foreign-key-aware insertion order
  3. 03Deterministic sample values
  4. 04SQL or JSON output
  5. 05Reports parser confidence and skipped tables

Use cases

  1. 01Create repeatable local development fixtures
  2. 02Prepare seed data for automated tests
  3. 03Generate dependency-ordered inserts from a schema

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/hainrixz/claude-db --skill seed

After installing, type/seedin Claude Code to confirm it loaded.

Details

From the project README

/claude-db:seed

Generates FK-aware, deterministic sample/seed INSERTs from a schema. Read-only — it never connects to or writes to a database. The output is dev/test data only: predictable values derived from the row index (no RNG), suitable for fixtures, local development, and CI — never for production.

$ARGUMENTS = <path-to-schema> [flags]. The target is a schema/DDL, ORM model, or migration file. If no path is given, detect one in the working directory (look for *.sql, schema.prisma, models.py, migration dirs). If nothing is found, say so plainly and suggest pointing at a file or running /claude-db:design to create one.

What to do

  1. Resolve the schema path (from $ARGUMENTS or detection above).
  2. Run the generator: 
    node scripts/gen-seed.mjs --file <schema> --rows N [--format sql|json]
    • --rows N defaults to 5 (clamped 1–1000). --format defaults to sql; json returns the statements as an array plus metadata.
    • The script topologically sorts tables so each table's FK parents are inserted first, fills FK columns with valid parent ids, and skips auto-increment/identity PKs.
  3. Present the result:
    • The insert order (dependency order, parents → children) and per-table row counts.
    • The generated INSERT statements, grouped by table in that order.
    • The parser confidence and any tables skipped or cycles ignored.
  4. Offer to save to a file (e.g. seed.sql or fixtures.json) — only write if the user confirms a path. Otherwise leave it inline.

Notes

  • Deterministic by design: re-running with the same --rows yields byte-identical output, so seeds are diff-friendly and reproducible.
  • Values are illustrative placeholders (emails like [email protected], fixed timestamps, sequential ids) — not realistic or privacy-safe production data.
  • Cyclic FKs are ignored for ordering; deferred-constraint or self-referential cases may need a manual pass after generation.

For dev/test fixtures only. Never run generated seeds against a production database. Respond in the user's language (EN/ES).

Comments

Markdown supported