How do I install the Perf Audit skill?

Run npx skillmds@latest add lucassantana-dev/perf-audit in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Perf Audit safe to use?

SkillMD's automated safety review verdict for this skill is PASS. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. Capability flags: makes network calls. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Perf Audit?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Perf Audit free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Perf Audit?

LucasSantana-Dev (@lucassantana-dev) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Perf Audit by lucassantana-dev — Agent Skill

Name: Perf Audit
Rating: 4.666666666666667 (3 reviews)
Author: LucasSantana-Dev

Perf Audit

End-to-end performance audit composite — baseline the slow metric, profile to find the real hotspot, implement the smallest targeted fix, re-benchmark to confirm improvement, capture in ADR. Use when users report slowness, CI shows runtime regression, or before a release. Never optimize without profiling first.

LucasSantana-Dev Updated Aug 19, 2026 1 repo stars

File contents

perf-audit

Profile before you optimize. Never guess hotspots.

Phases

Phase 1 — Baseline

Measure current performance and record the numbers (test suite runtime, build time, bundle size, API latency, DB query time). These go into the reconciliation block as before-state.

Phase 2 — Profile (mandatory before fixing)

Locate the top 1-3 hotspots using the appropriate tool for the stack (flame graph, bundle analyzer, EXPLAIN ANALYZE, verbose test reporter). Do NOT skip this phase and jump to a fix based on intuition.

Phase 3 — Fix (smallest change, biggest hotspot)

Apply one fix per PR. If the fix would touch >5 files or >150 LOC → stop and escalate to /research-and-decide.

Common patterns: memoize pure functions in hot loops, lazy-load eager imports, add DB indexes, replace sync I/O with async, deduplicate test setup.

Phase 4 — Validate

Re-run the same benchmarks from Phase 1 under the same conditions. Confirm improvement is >5% (above the noise floor). If <5% → the fix may not be worth the change risk; revert and document.

Phase 5 — Capture

Write an ADR with context (which metric, by how much), decision, before/after numbers, and revisit trigger.

Reconciliation

PERF AUDIT — <scope>

Phase 1 Baseline:  <metric>: <before value> ✅
Phase 2 Profile:   Hotspot: <file/function/query> ✅
Phase 3 Fix:       <change applied> | (escalated to /research-and-decide) ✅
Phase 4 Validate:  <metric>: <before> → <after> (<delta>%) ✅
Phase 5 Capture:   ADR <path> ✅ | (skipped: below noise floor)

Failure / Stop Conditions

Phase 2: profiling tools unavailable → baseline with available metrics, note gap in ADR, continue.
Phase 3: fix scope exceeds 5 files / 150 LOC → stop, escalate to /research-and-decide.
Phase 4: no measurable improvement → revert, mark "investigated, no action" in ADR, stop.
Never ship a fix whose only evidence is subjective. Require numbers.

---name: perf-auditdescription: End-to-end performance audit composite — baseline the slow metric, profile to find the real hotspot, implement the smallest targeted fix, re-benchmark to confirm improvement, capture in ADR. Use when users report slowness, CI shows runtime regression, or before a release. Never optimize without profiling first.---# perf-auditProfile before you optimize. Never guess hotspots.## Phases### Phase 1 — BaselineMeasure current performance and record the numbers (test suite runtime, build time, bundle size, API latency, DB query time). These go into the reconciliation block as before-state.### Phase 2 — Profile (mandatory before fixing)Locate the top 1-3 hotspots using the appropriate tool for the stack (flame graph, bundle analyzer, `EXPLAIN ANALYZE`, verbose test reporter). Do NOT skip this phase and jump to a fix based on intuition.### Phase 3 — Fix (smallest change, biggest hotspot)Apply one fix per PR. If the fix would touch >5 files or >150 LOC → stop and escalate to `/research-and-decide`.Common patterns: memoize pure functions in hot loops, lazy-load eager imports, add DB indexes, replace sync I/O with async, deduplicate test setup.### Phase 4 — ValidateRe-run the same benchmarks from Phase 1 under the same conditions. Confirm improvement is >5% (above the noise floor). If <5% → the fix may not be worth the change risk; revert and document.### Phase 5 — CaptureWrite an ADR with context (which metric, by how much), decision, before/after numbers, and revisit trigger.## Reconciliation```PERF AUDIT — <scope>Phase 1 Baseline: <metric>: <before value> ✅Phase 2 Profile: Hotspot: <file/function/query> ✅Phase 3 Fix: <change applied> | (escalated to /research-and-decide) ✅Phase 4 Validate: <metric>: <before> → <after> (<delta>%) ✅Phase 5 Capture: ADR <path> ✅ | (skipped: below noise floor)```## Failure / Stop Conditions- Phase 2: profiling tools unavailable → baseline with available metrics, note gap in ADR, continue.- Phase 3: fix scope exceeds 5 files / 150 LOC → stop, escalate to `/research-and-decide`.- Phase 4: no measurable improvement → revert, mark "investigated, no action" in ADR, stop.- Never ship a fix whose only evidence is subjective. Require numbers.

Frequently asked questions