# Validate Objects

> Validate Objects

- Skill: `snowflake-labs/validate-objects` (Agent Skill, multi-file: 5 files)
- Install (CLI): `npx skillmds@latest add snowflake-labs/validate-objects`
- Raw SKILL.md: https://api.skillmd.com/api/skills/snowflake-labs/validate-objects/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: snowflake-labs (https://skillmd.com/u/snowflake-labs)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/snowflake-labs/validate-objects

---


# Validate Objects

## On Entry

Tell the user:
> **Data Validation** — I'll compare your migrated data against the source to verify row counts, schema matches, and data integrity.

## Step 1: Configure Session

Call `configure()` to retrieve the current configuration.

Check the configured source dialect first.

- **Snowflake source** → require `snowflake_connection` and
  `snowflake_database`; no separate `source_connection` or DEW is required.
  Confirm the connection/database with the user.
- **Other sources** → require `snowflake_connection`, `source_connection`, and
  `snowflake_database`. Confirm all three; collect and configure any missing
  values.

Use the shared infrastructure configured during setup. Do not ask local vs
SPCS again, choose a pool, or start/repair infrastructure from a per-object
validation task. If dispatch reports that infrastructure is unavailable,
return the remediation to the main agent; it resumes the saved placement
through [`../data-infrastructure/SKILL.md`](../data-infrastructure/SKILL.md).
Only an explicit user request changes that persisted setup.

## Step 2: Validate

Load [actions/validate_tables.md](actions/validate_tables.md). Snowflake-source
projects generate a template without a registry filter; the user or agent must
fill its source and target object names before the run.

## Step 3: Wave progress

The **error-first data validation report** (Result + Workflow, Errors,
Suggested fixes) is produced in
[actions/validate_tables.md](actions/validate_tables.md) **Step 5** from
Monitor's terminal `detail.completion`. Use `job_status(details=true)` only
for deeper failure diagnostics; do not substitute `migration_status()`.

After `validate_tables.md` completes, optionally call `migration_status()` for wave-level context only:

- If the wave is complete, the next `configure()` call will auto-advance to the next wave.
- A per-object subagent does not tear down shared infrastructure. The main
  agent owns end-of-wave/session teardown after all active slots finish.