har-analysis

name	har-analysis
description	Parse HAR files and analyze web performance — waterfall, slow requests, caching issues, optimization opportunities
license	MIT
author	video-editor
version	1.0.0
user_invocable	true

HAR Analysis

Parse HTTP Archive (HAR) files and produce actionable performance insights using the visualize skill for charts and diagrams.

When to Apply

• User says "analyze this HAR", "speed this up", "why is this slow", "performance", or provides a .har file path
• User asks about load times, request waterfalls, caching headers, or transfer sizes

Procedure

1. Locate the HAR file: User provides a path (often in ~/Downloads/). Verify it exists.

2. Run the analysis script:

   /c/Python314/python scripts/har-analysis.py "<path-to-har-file>" --output "$TEMP/har-analysis.json" 2>&1
   

   The script writes two files:

   • $TEMP/har-analysis.json — structured analysis results
   • $TEMP/viz-data.json — visualization panels for the visualize skill

3. Read the analysis JSON and present findings to the user:

   • Top slowest requests with timing breakdown (wait vs receive vs blocked)
   • Caching issues (missing Cache-Control, no-store on static assets, missing ETags)
   • Compression gaps (large uncompressed responses)
   • Request waterfall showing parallelism and blocking chains
   • Transfer size breakdown by content type

4. Launch visualization (uses the visualize skill's infrastructure):

   cd c:/Users/imank/projects/video-editor && python scripts/visualize.py > /dev/null 2>&1 & disown
   

5. Recommend optimizations based on findings — prioritized by impact.

Analysis Categories

Category	What It Checks
Slow Requests	Requests >200ms, broken down by wait/receive/blocked/DNS/SSL
Caching	Missing or weak Cache-Control, no ETag/Last-Modified, no-store on cacheable resources
Compression	Responses >1KB without Content-Encoding (gzip/br/zstd)
Waterfall	Sequential chains that could be parallelized, long gaps between requests
Size	Large responses, breakdown by content type
CORS	Preflight (OPTIONS) overhead, count and time spent
Redirects	3xx chains that add latency
Errors	4xx/5xx responses

What the Script Outputs

The har-analysis.json contains:

{
  "summary": { "total_requests", "total_size_kb", "total_time_ms", "page_load_time_ms" },
  "slow_requests": [...],
  "caching_issues": [...],
  "compression_gaps": [...],
  "waterfall": [...],
  "by_content_type": {...},
  "cors_overhead": {...},
  "recommendations": [...]
}

Without the Script

If the script isn't available, Claude can run inline Python to parse the HAR:

/c/Python314/python -c "
import json
with open('<path>', 'r', encoding='utf-8') as f:
    har = json.load(f)
# ... analysis code ...
"

This is acceptable for quick one-off analysis but the script is preferred for consistent output.