gtsi-code-review

Reviews diffs and pull requests against operational governance policies and issues a merge verdict.

Updated May 21, 2026
One-click install
npx skills add https://github.com/gestaosian/gtsi-claude-skills --skill gtsi-code-review-gestaosian
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: gtsi-code-review
Source: https://github.com/gestaosian/gtsi-claude-skills/tree/main/plugins/gtsi-ops-plugin/skills/gtsi-code-review
Command: npx skills add https://github.com/gestaosian/gtsi-claude-skills --skill gtsi-code-review-gestaosian

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Reviewing pull requests against internal data governance rules, SIAN platform patterns, and operational risk criteria is inconsistent when done manually, and violations like direct data corrections in Silver/Gold layers can slip into production. ## Core Features & Use Cases - Five review lenses: Checks data correction policy violations, SIAN DAG factory patterns and naming conventions, governance (hardcoded credentials, Airflow Variables), operational risks (rollback plans, destructive changes), and general code quality. - Structured verdicts: Issues approve, approve-with-caveats, or refuse verdicts with blocker/attention/observation findings, and persists the report to .gtsi/reviews/<author>/ with automatic author detection from git config. - Use Case: Before merging a PR that adds a new Airflow DAG, ask for a review; the skill fetches the diff via gh pr view or git diff, flags a missing cluster_by for an Iceberg table and a typo in a log message, and returns an approve-with-caveats verdict. ## Quick Start Ask the assistant to review PR #42 or the current branch diff before merge and give a verdict.

Frequently Asked Questions about gtsi-code-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review a pull request before merging?▼

Provide the PR number or run git diff against the main branch, then ask for a code review. The skill reads the full diff, applies five review lenses, and returns an approve, approve-with-caveats, or refuse verdict with specific findings.

What does an operational code review check in Airflow DAGs?▼

It verifies DAG factory pattern compliance, naming convention {sistema}__{tabela}__{cliente}__{frequência}.py, correct sys.path depth, Iceberg cluster_by configuration, and that new DAGs are created paused by default.

Can a code review reject changes that modify data directly?▼

Yes. Direct UPDATE/DELETE on Silver or Gold tables, dbt CASE WHEN overrides to skip records, and correction scripts in derived layers violate the data correction policy and result in a refuse verdict, redirecting fixes to the source system.

Where are code review reports saved?▼

Reports are saved to .gtsi/reviews/<author-slug>/<date>-<pr-or-branch>.md, with the author detected automatically from git config user.name. The verdict and findings are also shown inline in the conversation.

What are the limitations of automated PR review?▼

The review requires an actual diff or PR number and will not proceed without one. It evaluates against documented policies and patterns but does not execute code or run tests, so runtime behavior still needs separate validation.