- Home
- Skills
- Ratacat
- Claude Skills
- Deployment Verification Agent
deployment-verification-agent_skill
- Python
24
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill ratacat/claude-skills --skill deployment-verification-agent- SKILL.md5.6 KB
Overview
This skill produces executable Go/No-Go deployment checklists for any PR that touches production data, migrations, backfills, or data-processing logic. It creates concrete pre/post-deploy SQL checks, documents destructive/locking steps, defines rollback procedures, and outlines immediate and 24-hour monitoring. Use it to remove guesswork and enable fast, confident launch decisions.
How this skill works
Given a PR description or diff, the agent identifies critical data invariants, generates read-only SQL queries for baseline and post-deploy verification, and translates migration/backfill steps into an actionable runbook with estimated runtime and batching. It also produces rollback steps and a monitoring plan with metrics, alert thresholds, and console spot checks. The output is a one-page checklist engineers can execute during deployment.
When to use it
- PR adds or changes database migrations that modify data
- PR changes data-processing or classification logic (possible duplicates/loss)
- PR includes backfills, batch jobs, or irreversible denormalization
- Before enabling feature flags that alter production records
- When you need a clear Go/No-Go decision and measurable verification
Best practices
- Define explicit data invariants and acceptance tolerances before running any changes
- Capture baseline results from pre-deploy SQL and store them with the deploy ticket
- Run read-only verification queries within 5 minutes of deploy and compare to baseline
- Use batching, retries, and idempotent backfills to reduce blast radius
- Prepare a tested rollback path and verify backups exist before starting
Example use cases
- PR changing email classification: generate invariants, pre/post SQL to compare classification counts, and rollback steps to restore prior labels
- Database migration that adds a new column + backfill: provide batching plan, estimated runtimes, and post-deploy NULL checks
- Backfill of user status: create baseline counts, sampling queries, and monitoring alerts for unexpected status changes
- Feature toggle enabling new write path: checklist for dual-write validation, migration finish checks, and toggle rollback steps
FAQ
Document why irreversible changes are acceptable, ensure a recent backup exists, and create manual restoration steps. Mark the checklist with clear risk tolerances.
How do I choose alert thresholds?
Use baseline variance: set alerts for deviations beyond expected tolerances (e.g., error rate >1% or missing-data count >0) and tune after initial monitoring window.