# Audit Performance

Audit and prioritize performance work through bounded-work and forward-progress checks.

- Repository: https://github.com/dzhng/skills (1394 stars)
- Category: Engineering
- License: MIT
- Last updated: 2026-11-01
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add dzhng/skills --skill audit-performance -g -a claude-code -y`
Codex: `npx skills add dzhng/skills --skill audit-performance -g -a codex -y`
Cursor: `npx skills add dzhng/skills --skill audit-performance -g -a cursor -y`
Gemini CLI: `npx skills add dzhng/skills --skill audit-performance -g -a gemini-cli -y`

## Description

Audit and prioritize performance work through bounded-work and forward-progress checks. Use when asked to find performance issues, investigate freezes/500s/OOMs, review retries or polling for no-progress loops, rank a performance backlog, or implement the next simple performance fixes.

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# Audit Performance

Find work that grows without a useful bound or repeats without progress. Produce an evidence-backed
ledger or report, then prefer the smallest fix that preserves recovery.

## Workflow

1. **Trace hot paths.** Find recurring syncs, polls, streams, retries, queues, scheduled jobs,
   filesystem walks, and request-time reads. Follow each through its production caller; ignore
   test-only paths.
2. **Compute the amplification.** State the trigger and worst-case work in concrete units: rows,
   queries, pages, entries, bytes, jobs, retries, or full-file rewrites. Distinguish a bounded large
   constant from work that grows with history, tenants, paths, or outage duration.
3. **Check forward progress.** For every retry, fixed-prefix batch, cutoff, or transitional poll,
   identify what changes before the next attempt. If nothing must change, it can loop or starve
   forever. Check that partial fixes do not merely delay the same work.
4. **Falsify severity.** Inspect existing caps, indexes, backoff, deadlines, healing owners, and
   caller frequency. Dismiss findings already bounded cheaply enough or continuously healing.
5. **Record before fixing.** If the project keeps a performance ledger, append every supported
   finding and important dismissal using its existing convention. Otherwise return a compact report.
   Include priority, trigger, impact, current owner, acceptance seam, and evidence.
6. **Prioritize with the policy below.** If implementation is authorized, invoke the project's
   test-writing workflow and prove the old behavior red at the outermost practical seam. Fix, review,
   and update the ledger or report with verification evidence.

## Priority Policy

Rank highest when the issue can cause a user-visible freeze or error, memory/disk growth, blocked
request processing, data loss, fleet-wide amplification, or a poison item that prevents later work.

The rest: https://github.com/dzhng/skills/blob/HEAD/skills/engineering/audit-performance/SKILL.md