agentleFS
Sign inStart free
ChariereFiedlerChariereFiedler/ronce-racine/skills/audit-performance-frontend/SKILL.md

audit-performance-frontend

Skill9 starsChanged 23 days agoMITView on GitHub

Low-level frontend performance audit, specialized for WebKit/JSC and embedded or real-time contexts - on a mainstream stack (React/Chrome SPA) many questions are N/A and the grid adapts. Use when auditing frontend rendering performance, WebKit internals, JSC optimization, DOM structure, CSS rendering pipeline, memory/GC, cache strategy, or profiling methodology. Triggers on "audit performance", "audit frontend", "audit perf", "audit rendering", "audit webkit", "audit jsc".

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    ME when: an audit scoped to this domain only

    line 12
  2. Distinctive:

    audit-industrialisation: instead if: a cross-domain global audit - it orchestrates every audit (including ME) and produces a consolidated …

    line 13
  3. Very specific:

    Category 3 - JS Engine: look for polymorphism patterns, delete, arrow functions in loops, eval, sparse arrays.

    line 56
  4. Distinctive:

    Category 4 - Memory: check allocations in loops, event listeners without cleanup, caches without eviction.

    line 57
  5. Distinctive:

    Category 6 - Profiling: check instrumentation, metrics, performance CI.

    line 59
  6. Distinctive:

    Every question applies with the standard grid

    line 80

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

When ME and not audit-industrialisation

  • ME when: an audit scoped to this domain only
  • audit-industrialisation instead if: a cross-domain global audit - it orchestrates every audit (including ME) and produces a consolidated report

For a full project audit, use audit-industrialisation rather than invoking each skill separately.

Low-Level Frontend Performance Audit

Scope check before starting: this grid targets WebKit/JSC internals and embedded/real-time constraints. On a mainstream stack (e.g. a React SPA on evergreen browsers), mark the engine-specific questions N/A per the Variants section instead of forcing answers - a mostly-N/A domain is a valid outcome.

Overview

Audit a project's frontend performance maturity, targeting sub-100ms rendering. Covers the integrity of the rendering pipeline: DOM structure, CSS rendering, JS engine internals (JSC), memory and GC management, network/cache strategy, and profiling tooling. Micro-optimization oriented, for embedded or real-time contexts running on WebKit.

29 questions split across 6 categories:

CategoryQuestionsGrid
1 - HTML & Critical Path4 (PF-01 to PF-04)reference/html-critical-path.md
2 - CSS & Rendering Pipeline5 (PF-05 to PF-09)reference/css-rendering-pipeline.md
3 - JavaScript Engine & Runtime JSC9 (PF-10 to PF-18)reference/js-engine-runtime.md
4 - Memory & Garbage Collection4 (PF-19 to PF-22)reference/memory-gc.md
5 - Assets, Cache & Network4 (PF-23 to PF-26)reference/assets-cache-network.md
6 - Profiling & Tooling3 (PF-27 to PF-29)reference/profiling-tooling.md

Scoring

  • Each scored question gets a level from 0 to 4.
  • Questions marked N/A are excluded from the calculation.
  • Domain score = sum of levels / number of scored questions (excluding N/A).
  • Criticality (used to prioritize remediation, NOT in the score - see audit-report/reference/scoring-model.md):
    • MUST: PF-01, PF-02, PF-03, PF-06, PF-10, PF-13, PF-17, PF-23, PF-27, PF-28, PF-29
    • SHOULD: PF-04, PF-05, PF-07, PF-09, PF-11, PF-12, PF-14, PF-15, PF-19, PF-20, PF-21, PF-22, PF-24, PF-25, PF-26
    • COULD: PF-08, PF-16, PF-18

Audit protocol

  1. Detect the project type: SPA, SSR, PWA, embedded WebKit, component library.
  2. Category 1 - HTML: scan for document.write, measure DOM depth, check blocking scripts and priority attributes.
  3. Category 2 - CSS: audit selectors, animations, compositing, custom properties, fonts.
  4. Category 3 - JS Engine: look for polymorphism patterns, delete, arrow functions in loops, eval, sparse arrays.
  5. Category 4 - Memory: check allocations in loops, event listeners without cleanup, caches without eviction.
  6. Category 5 - Assets: measure critical-path size, check compression, cache headers, font loading.
  7. Category 6 - Profiling: check instrumentation, metrics, performance CI.
  8. Run the verification commands for each question (see grids).
  9. Assign a level (0-4) per question with justification and confidence level.
  10. Mark N/A the non-applicable questions (e.g. PF-10 to PF-18 if there is no custom JS, just a framework).
  11. Produce the report (output format below).

Variants by project type

SSR application (Nuxt, Next.js)

  • PF-01 to PF-04: Fully applicable, the server generates the initial HTML
  • PF-23: Adapted - TTFB includes server render time

UI component library

  • PF-23 to PF-26: N/A (no control over deployment)
  • PF-27 to PF-29: Adapted - component performance benchmarks

Static site (11ty, Hugo, Astro)

  • PF-10 to PF-18: Often N/A (little custom JS)
  • PF-01 to PF-09, PF-23 to PF-29: Fully applicable

Embedded WebKit application (WKWebView, Electron)

  • Every question applies with the standard grid
  • Particular focus on PF-10 to PF-22 (memory/CPU constraints)

Grids by category

Each grid contains, per question: statement, criticality, elements to analyze, bash verification commands, points to check, and a 0-4 level table.

  • reference/html-critical-path.md - Category 1, HTML & Critical Path, 4 questions (PF-01 to PF-04)
  • reference/css-rendering-pipeline.md - Category 2, CSS & Rendering Pipeline, 5 questions (PF-05 to PF-09)
  • reference/js-engine-runtime.md - Category 3, JavaScript Engine & Runtime JSC, 9 questions (PF-10 to PF-18)
  • reference/memory-gc.md - Category 4, Memory & Garbage Collection, 4 questions (PF-19 to PF-22)
  • reference/assets-cache-network.md - Category 5, Assets, Cache & Network, 4 questions (PF-23 to PF-26)
  • reference/profiling-tooling.md - Category 6, Profiling & Tooling, 3 questions (PF-27 to PF-29)

Output format

## Frontend Performance - Global score: X.X/4 (Y questions scored out of 29)

### Summary
[2-3 sentences summarizing frontend performance maturity]
[Detected project type: SPA / SSR / PWA / embedded / static]

### Detail by question

#### Category 1 - HTML & Critical Path

| Code | Question | Criticality | Level | Confidence | Justification |
|------|----------|-----------|--------|-----------|---------------|
| PF-01 | DOM structure and parsing | MUST | X | high/medium/low | ... |
| PF-02 | Render-blocking | MUST | X | high/medium/low | ... |
| PF-03 | Layout thrashing | MUST | X | high/medium/low | ... |
| PF-04 | Loading & priority | SHOULD | X | high/medium/low | ... |

#### Category 2 - CSS & Rendering Pipeline

| Code | Question | Criticality | Level | Confidence | Justification |
|------|----------|-----------|--------|-----------|---------------|
| PF-05 | CSS selectors | SHOULD | X | high/medium/low | ... |
| PF-06 | Layout triggers | MUST | X | high/medium/low | ... |
| PF-07 | Paint & GPU | SHOULD | X | high/medium/low | ... |
| PF-08 | Custom properties | COULD | X | high/medium/low | ... |
| PF-09 | Font rendering | SHOULD | X | high/medium/low | ... |

#### Category 3 - JavaScript Engine & Runtime

| Code | Question | Criticality | Level | Confidence | Justification |
|------|----------|-----------|--------|-----------|---------------|
| PF-10 | JSC monomorphism | MUST | X | high/medium/low | ... |
| PF-11 | NaN-boxing | SHOULD | X | high/medium/low | ... |
| PF-12 | JSC compilation tiers | SHOULD | X | high/medium/low | ... |
| PF-13 | Deoptimization | MUST | X | high/medium/low | ... |
| PF-14 | Functions & closures | SHOULD | X | high/medium/low | ... |
| PF-15 | JSC arrays | SHOULD | X | high/medium/low | ... |
| PF-16 | Strings & RegExp | COULD | X | high/medium/low | ... |
| PF-17 | Event loop | MUST | X | high/medium/low | ... |
| PF-18 | Weak references | COULD | X | high/medium/low | ... |

#### Category 4 - Memory & GC

| Code | Question | Criticality | Level | Confidence | Justification |
|------|----------|-----------|--------|-----------|---------------|
| PF-19 | GC Riptide | SHOULD | X | high/medium/low | ... |
| PF-20 | Allocation pressure | SHOULD | X | high/medium/low | ... |
| PF-21 | Memory leaks | SHOULD | X | high/medium/low | ... |
| PF-22 | GPU memory | SHOULD | X | high/medium/low | ... |

#### Category 5 - Assets, Cache & Network

| Code | Question | Criticality | Level | Confidence | Justification |
|------|----------|-----------|--------|-----------|---------------|
| PF-23 | Network critical path | MUST | X | high/medium/low | ... |
| PF-24 | Compression & encoding | SHOULD | X | high/medium/low | ... |
| PF-25 | Multi-layer cache | SHOULD | X | high/medium/low | ... |
| PF-26 | Font loading | SHOULD | X | high/medium/low | ... |

#### Category 6 - Profiling & Tooling

| Code | Question | Criticality | Level | Confidence | Justification |
|------|----------|-----------|--------|-----------|---------------|
| PF-27 | Methodical profiling | MUST | X | high/medium/low | ... |
| PF-28 | Sub-100ms metrics | MUST | X | high/medium/low | ... |
| PF-29 | Performance regression CI | MUST | X | high/medium/low | ... |

### Strengths
- ...

### Weaknesses
- ...

### Recommendations
1. [MUST] ...
2. [SHOULD] ...
3. [COULD] ...

### Non-auditable items
- [List the items impossible to verify by static analysis: runtime metrics, JIT profiling, real GC pauses, real-world VRAM, etc.]

Exit condition

  • Project type detected and matching variant applied
  • All 29 questions reviewed (scored 0-4 or marked N/A with justification)
  • Verification commands run for each scored question
  • Confidence level assigned per question
  • Global score computed (excluding N/A)
  • Report produced in the output format, with strengths/weaknesses, prioritized recommendations, and non-auditable items

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free