debug-prod-issues_skill

Overview

This skill helps debug production issues for a TypeScript, CLI-first banking agent deployed on Vercel and using a Neon Postgres production database. It provides concrete commands and scripts to view and filter Vercel function logs, wait for deployments reliably, connect to the production database, and inspect key tables. Use it to quickly triage API errors, data discrepancies, and deployment problems.

How this skill works

It inspects Vercel deployments and function logs (with --scope prologe) and offers patterns to stream, filter, and timebox log queries to avoid hangs. It guides safe connections to the Neon Postgres production database by loading .env.production.local before importing DB modules and provides example scripts to query important tables. It also enforces robust deployment waiting using vercel inspect --wait instead of blind sleep.

When to use it

• Investigating production API endpoint errors and stack traces
• Confirming whether a deployment is complete and live
• Streaming and filtering serverless function logs to surface errors
• Running one-off database scripts against production data safely
• Verifying environment variables and their propagation after changes

Best practices

• Always run Vercel commands with --scope prologe for this project to avoid wrong-team operations
• Use vercel inspect --wait --timeout instead of sleep to detect real deployment status
• Load .env.production.local before importing or initializing the DB in scripts to ensure correct POSTGRES_URL
• Timebox long vercel logs with timeout or pipe to head to avoid hangs
• Prefer dynamic imports of database modules in debugging scripts so dotenv runs first

Example use cases

• Triage an "Internal server error" by streaming recent Vercel logs, grepping for stack traces, then curl-ing the endpoint to reproduce
• After pushing a fix, get the latest deployment URL, run vercel inspect --wait to confirm Ready status, then test the live route
• Connect to Neon production, run a one-off script that counts records in user_safes and incoming_deposits to validate syncs from Safe API
• Investigate a missing Safe for a workspace by querying user_safes and users to compare workspaceId and primaryWorkspaceId
• Check that an env var is set in production with vercel env ls --scope prologe and redeploy if necessary