# Submission Guide

## What To Review First

1. [README.md](README.md) - setup, scope, and run commands.
2. [sample_report/20260617T153847Z-neemans-com.html](sample_report/20260617T153847Z-neemans-com.html) - tracked Neeman's proof report.
3. [AD_CANDIDATE_AUDITS.md](AD_CANDIDATE_AUDITS.md) - latest Neeman's and Brooklinen live-run screenshots.
4. [ARCHITECTURE.md](ARCHITECTURE.md) - data flow and design.
5. [COMPLETION_AUDIT.md](COMPLETION_AUDIT.md) - requirement-to-evidence checklist.

## Primary Demo Command

```bash
python run_audit.py --url "https://neemans.com/products/duoform-sneakers-for-men-ecru"
```

Optional:

```bash
GROQ_API_KEY=... python run_audit.py \
  --url "https://neemans.com/products/duoform-sneakers-for-men-ecru" \
  --llm-duplicates
```

## Why This Submission Is Assignment-Aligned

- Uses a real browser rather than static scraping.
- Captures reviewer-inspectable artifacts: screenshots, HAR, trace, raw audit JSON, network JSONL, HTML report, JSON summary.
- Proves the required Neeman's ecommerce path.
- Separates universal URL checks from deep ecommerce checks.
- Uses deterministic rules for report generation.
- Keeps LLM usage optional, narrow, and logged.
- Includes negative evidence from other pages to show the tool catches unsafe ad spend instead of rubber-stamping every page.

## Video Status

No demo video file is included. The repo includes live-run screenshots and browser-generated HTML reports. Playwright traces are produced under `artifacts/<run_id>/trace.zip` for local inspection after a run.