neon-postgres-egress-optimizer

Postgres Egress Optimizer is a diagnostic skill for finding and fixing application-side query patterns that cause excessive data transfer (egress) from a Postgres database, which is a common driver of surprise database bills. Its premise is that most high egress comes from applications fetching more data than they use, so it guides the user through a disciplined four-step method: diagnose which queries transfer the most data, analyze the codebase behind them, fix the anti-patterns, and verify that nothing broke and the transfer actually dropped. It is a child of the parent neon skill, which it points the agent to first.

For diagnosis it uses the read-only pg_stat_statements extension, with queries that rank statements by total rows returned, rows per execution, call frequency, and total execution time, plus guidance on handling empty stats after a Neon compute scales to zero. It teaches how to interpret results by combining row counts with column width (JSONB, TEXT, BYTEA) to estimate real egress impact. The fix section catalogs the common anti-patterns — SELECT * fetching unused wide columns, missing pagination on list endpoints, high-frequency queries on static data that should be cached, and application-side aggregation that transfers whole tables to compute a small summary — each with a before/after example and, where relevant, pushing work into SQL. A verify step confirms correctness and measures the reduction.

It is aimed at application developers, backend engineers, and cost-conscious teams who see egress spikes, high Neon bills, or want to review query patterns for cost efficiency. The main prerequisite is access to run diagnostic SQL (ideally against production stats) and the codebase to analyze.