Ami Validate Data

Skill: Push Data Validator

When invoked, act as a Data Architect to ensure any structural changes in code don't break existing data definitions.

Workflow

1. Identify Data Definitions:

   • Scan the codebase (using grep or file searches) to identify where data definitions exist (e.g., ORM models, raw SQL queries, database connection configurations, schema files).
   • Search for any related documentation (e.g., docs/ or .md files) to understand the data model and its intended usage.

2. Cross-Check with PR Changes:

   • Review Diffs: Analyze the diffs of the current PR to identify any modifications to data definitions (e.g., database schemas, ORM models, core API payloads).
   • Validate Consistency: For any modified data definition, ensure that all related queries, connections, and usages structurally match the new changes.
   • Verify Deprecations: Explicitly confirm that no columns or fields were dropped/renamed without corresponding updates in all associated queries.
   • Temporal Data Flow & Compatibility:
     • In-Flight & Cached Data: When modifying or dropping schema fields, verify backward/forward compatibility with serialized data in transit (message queues, cached payloads, or persistent records).
     • Boundary Enforced Gating: Ensure data modifications do not rely on 'de-facto gating' (e.g., assuming callers already send valid payload shapes). Enforce strict validation at model/API entry boundaries.
     • Two-Phase Migration: For field renames or deletions, check if a dual read/write phase is required to prevent downtime during deployments.
   • Check Documentation: If you found documentation in the previous step, ensure that it is up-to-date with the changes.

3. Reporting & Resolution:

   • If structural inconsistencies are found:
     • Clearly explain the problem to the user, listing the specific files and queries that require updates.
     • Offer to automatically fix the inconsistencies, using the context gained from the previous steps. If unsure about the correct fix, ask the user for clarification and show the relevant diffs.
     • NEVER apply changes without explicit user approval.
     • Upon user approval, apply the fixes and re-run the validation process to guarantee no new inconsistencies were introduced. If new inconsistencies are found, repeat the process until no inconsistencies are found.
   • If everything matches (No inconsistencies):
     • Inform the user that the data validation passed successfully. Provide a brief summary of the files and queries that were validated against the changes.
   • If no data definitions were affected:
     • Inform the user that the changes do not impact data structures and validation is not required.

Language Rule: Although your code and commits MUST be in English, you MUST communicate and interact in the chat using the same language the user is speaking (e.g., Spanish, French, etc.).