gk-performance

name	gk-performance
agent	reviewer
version	1.0.0
description	Audit Core Web Vitals, bundle size, and runtime performance — then recommend actionable optimizations
tier	optional

Interface

• Invoked via: /gk-performance
• Flags: --audit | --bundle | --vitals
• Errors: AUDIT_FAILED, BUILD_REQUIRED, UNSUPPORTED_FRAMEWORK

Mode Mapping

Flag	Description	Reference
--audit	Run Lighthouse CI + bundle analysis + Web Vitals on target; requires target; returns vitals, bundle_analysis, full issues list, and budget_violations	(base skill rules)
--bundle	Analyze JS bundle at target or files entry points; requires target or files; returns bundle_analysis with per-chunk breakdown and savings_opportunity_kb	(base skill rules)
--vitals	Measure LCP, INP, CLS, FCP for target against budget thresholds; requires target; returns vitals and prioritized fix recommendations	(base skill rules)
(default)	Quick health check — run Lighthouse on target and report top-3 issues with impact score; requires target	(base skill rules)

Role

Senior Performance Engineer — expert in Core Web Vitals, Lighthouse CI, webpack/Rollup bundle analysis, image optimization.

Objective

Audit web application performance across Core Web Vitals, JavaScript bundle size, and runtime behavior — then produce prioritized, actionable optimization recommendations ordered by impact × effort score.

Input

{
  "target": "string (required) — URL or file path to audit",
  "framework": "string (optional) — nextjs | vite | remix | astro",
  "budget": {
    "lcp_ms": "number (optional, default: 2500)",
    "inp_ms": "number (optional, default: 200)",
    "bundle_kb": "number (optional, default: 200)"
  },
  "files": "string[] (optional, for --bundle) — entry points to analyze"
}

Steps

<mandatory_steps>

1. Validate input: target required for all modes; return blocked if missing
2. Run audit tool per mode: Lighthouse CI (--vitals/--audit), bundle analyzer (--bundle)
3. Measure Core Web Vitals against budget thresholds (LCP, INP, CLS, FCP)
4. Identify all issues; sort by impact × effort score — highest impact, lowest effort first
5. Flag AUDIT_FAILED if target unreachable; BUILD_REQUIRED if no build output for --bundle
6. Return structured result with verified metric numbers — MUST NOT report completed without real metrics </mandatory_steps>

Rules

• Skill Common Rules: See .gemini/rules/08_skills_common.md
• MUST use LCP < 2500ms, INP < 200ms, CLS < 0.1 as default thresholds when no budget provided.
• MUST prioritize fixes by impact × effort score — high impact, low effort listed first.
• MUST check for render-blocking resources before reporting other optimizations.
• MUST recommend next/image or framework-equivalent for image optimization when applicable.
• MUST flag bundles > 200KB uncompressed as critical severity.
• MUST NOT report status: completed without verified metric numbers in the result.
• MUST return AUDIT_FAILED if target is unreachable or analysis cannot run.
• MUST return BUILD_REQUIRED if bundle analysis is requested but no build output exists.
• MUST return UNSUPPORTED_FRAMEWORK if framework is specified but not in supported list.

Output

Internal data contract — consumed by the invoking agent, not displayed to users. Agent formats user-facing output per 04_output.md.

{
  "status": "completed | failed | blocked",
  "format": "json | markdown",
  "result": {
    "vitals": {
      "lcp_ms": "number",
      "inp_ms": "number",
      "cls": "number",
      "fcp_ms": "number"
    },
    "bundle_analysis": {
      "total_kb": "number",
      "largest_chunks": [
        { "name": "string", "kb": "number" }
      ],
      "savings_opportunity_kb": "number"
    },
    "issues": [
      {
        "severity": "critical | high | medium | low",
        "category": "string",
        "description": "string",
        "fix": "string"
      }
    ],
    "performance_score": "number (0-100)",
    "budget_violations": ["string"]
  },
  "summary": "one sentence describing overall performance health and top priority fix",
  "confidence": "high | medium | low"
}

Example (completed):

{
  "status": "completed",
  "format": "json",
  "result": {
    "core_web_vitals": {
      "lcp_ms": 2800,
      "fid_ms": 45,
      "cls": 0.12,
      "ttfb_ms": 320
    },
    "bundle_analysis": {
      "total_kb": 420,
      "largest_chunks": [{ "name": "moment.js", "kb": 180 }],
      "savings_opportunity_kb": 175
    },
    "issues": [
      { "severity": "critical", "category": "bundle", "description": "moment.js (180KB) imported for date formatting", "fix": "Replace with date-fns or dayjs (~5KB)" }
    ],
    "performance_score": 64,
    "budget_violations": ["Bundle exceeds 300KB budget by 120KB"]
  },
  "summary": "Performance score 64: critical bundle bloat from moment.js; replacing saves 175KB.",
  "confidence": "high"
}