# Evidence
This folder holds the raw observations behind the report. Everything here is collected passively: ordinary page loads, the same as any visitor. Nothing is submitted, probed or enumerated.
## 1. Automated header and metadata check
```bash
npm run audit
```
Writes `passive-audit.md` (readable) and `passive-audit.json` (raw) here. It sends one GET per public page, 1.5 seconds apart, with a User-Agent that identifies it as an assessment script. It closes these report items:
| Finding | What to look at in `passive-audit.md` |
| --- | --- |
| S-06 | "Security headers" table: anything marked **missing** |
| S-07 | Not confirmable from outside; raise it as a config review item. A CSP with no `img-src` allow-list suggests image sources were never pinned |
| S-10 | "Well-known files": status of `security.txt` |
| P-01 | "Metadata" table: title in head, repeated brand, description, og:url |
| P-03 | "Redirects": apex to www should be a single 301/308 |
| P-04 | "Payload and forms": icon font column |
| P-05 | "Platform and caching": compare `/blog` with `/blog?category=…` cache status |
After running it, update the **Confidence** labels in `REPORT.md` from "To verify" to "Observed" (or remove a finding that turned out fine; that is a perfectly good result to report).
## 2. Browser DevTools checklist (5 minutes)
Use a normal browser session. Do not submit any form for testing.
1. **View source of `/careers/backend-developer-intern`** (Ctrl+U). Search for `
`: is it inside ``? Search `__next_f`: skim the serialized props for fields the page doesn't display, such as emails or internal IDs (S-08, P-01).
2. **Search the same source** for `turnstile`, `recaptcha`, `hcaptcha` (S-03).
3. **Network tab, reload `/contact`**: sort by size. Note the size of the Material Symbols font file and the total transferred (P-04).
4. **Network tab, click the first document request**: confirm the response headers match the script's output; screenshot them.
5. **Lighthouse (mobile, Performance + Best Practices + SEO)** on `/` and `/careers/backend-developer-intern`. Record LCP, TBT, CLS and the SEO score. One page load per run.
Save screenshots in `evidence/screenshots/` and reference them from the report if you like.
## 3. Optional third-party scanners
These tools load the homepage once from their own servers, like a visitor would:
- securityheaders.com: grade for response headers.
- developer.mozilla.org/en-US/observatory: headers plus cookie and redirect checks.
- pagespeed.web.dev: field and lab performance data.
Avoid active vulnerability scanners (Nikto, ZAP active scan, nuclei, directory brute-forcing). They are outside the permitted scope.
## 4. Observation log
Fill this in as you go; it makes the report easy to defend in an interview.
| Date | What I checked | How | Result | Updates finding |
| --- | --- | --- | --- | --- |
| 2026-11-01 | Page content and forms on 8 public pages | Page fetch | See REPORT.md | All "Observed" items |
| | Response headers | `npm run audit` | | S-06, P-05 |
| | RSC payload contents | View source | | S-08 |
| | Lighthouse scores | Chrome DevTools | | P-01 to P-04 |