Pentest Findings: What to Fix First
Fix penetration test findings in this order: first anything an outsider can exploit without logging in that touches customer data or admin access, then the rest by severity, with cheap quick wins fitted in along the way.
This post is for owners of small sites (WordPress, Shopify, indie SaaS, local business sites) who just received a pentest report and feel buried by it. You will get a severity-to-deadline table, a simple way to rank your penetration test findings, and exact steps to verify each fix and ask the tester for a retest.
The fix order in five steps
A pentest (penetration test) is an authorized, hands-on attempt to find weaknesses in your site. The report lists what was found. Work through it like this:
- Contain anything already being abused. If the report or your own logs show signs of a break-in, stop and follow a website hacked response plan first.
- Critical and high findings that anyone on the internet can reach. No login needed means no barrier for an attacker.
- Anything that exposes customer data or admin access, even if the label says Medium.
- Quick wins. Anything you can finish in under an hour, such as a header or a DNS record.
- Medium, then Low, then Informational. Batch these into your normal maintenance.
Severity-to-deadline table
These are common internal targets for small teams, not a legal standard. If a client contract or compliance framework sets stricter deadlines, follow those.
| Severity | What it usually means | Contain or reduce risk | Fully fixed by |
|---|---|---|---|
| Critical | Attacker can take over the site or read private data with little effort | Within 24 hours | 7 days |
| High | Serious flaw, may need a small condition to exploit | Within 3 days | 14 days |
| Medium | Needs special conditions or has limited impact | Plan within 14 days | 60 days |
| Low | Hardening or minor weakness | Add to backlog | 90 days or next maintenance window |
| Informational | No direct risk, good practice | Read and note | When convenient |
"Contain" means shrinking the danger before the real fix is ready. Examples: block a URL, turn off a plugin, or restrict a page to your office IP.
Why the severity label is only a starting point
Most testers rate a finding by two things: how likely an attacker is to use it, and how much damage results. OWASP's risk rating method is built on exactly that idea. Many reports also show a CVSS score, a 0 to 10 number from the FIRST CVSS standard that rates technical severity.
The label does not know your business. Adjust it with these questions:
- Can an anonymous visitor trigger it? If yes, move it up.
- Does it touch money, logins or personal data? If yes, move it up.
- Is it a known CVE (a public ID for a software flaw) that attackers are using right now? CISA keeps a list of these called the Known Exploited Vulnerabilities catalog. Search it for the CVE numbers in your report. A match goes to the top.
- Does it sit behind a login on an internal page only you use? If yes, you can usually move it down.
Group findings by root cause
Five findings are often one problem. An old WordPress plugin can create a version-disclosure finding, an XSS finding and a file-upload finding. Updating it closes all three. Group first, then count your real workload.
| Common finding | Typical fix | Effort |
|---|---|---|
| Outdated CMS, theme or plugin | Update, and delete anything unused | Minutes |
Exposed file such as .env, .git or a backup zip | Remove it or block access, then rotate every secret inside it | Minutes, plus rotation |
| Weak or default admin password | Set a long unique password and turn on MFA (multi-factor login) | Minutes |
| Missing security headers | Add response headers on the server or CDN | About an hour |
| Missing SPF or DMARC email records | Add DNS TXT records | About an hour, then monitoring |
| SQL injection or cross-site scripting (XSS) | Code change: parameterized queries, output encoding | Hours to days |
| Broken access control (users seeing other users' data) | Check permissions on the server for every request | Days |
The last two rows (injection, which includes SQL injection, and broken access control) are classic categories in the OWASP Top 10. XSS is covered by the injection category. They need a developer, so book that time early. The SPF and DMARC row is a DNS change, not a code change.
If a secret was exposed (a database password, an API key), assume someone copied it. Changing it is part of the fix. Deleting the file is not enough.
Quick wins with exact values
These are safe for most sites. Test on a staging copy first when you can.
Security headers. Headers are instructions your server sends to the browser. Here is a sensible starting set for Nginx:
add_header X-Content-Type-Options nosniff always;
add_header Referrer-Policy strict-origin-when-cross-origin always;
add_header Strict-Transport-Security "max-age=31536000" always;
Only add Strict-Transport-Security (HSTS, which forces HTTPS) once your whole site works over HTTPS. Be careful with includeSubDomains, since it also affects every subdomain. Do not paste a Content-Security-Policy from the internet. Start with Content-Security-Policy-Report-Only so nothing breaks, as explained in MDN's CSP guide. You can check your current headers with the free security headers tool.
Email spoofing records. If the report says anyone can send email pretending to be your domain, add a DMARC record. Create a TXT record at _dmarc.example.com with this value:
v=DMARC1; p=none; rua=mailto:dmarc-reports@example.com
p=none only watches and reports. After a few weeks of clean reports, and once SPF and DKIM are set up for every service that sends mail as your domain (newsletter tool, CRM, invoicing), move to p=quarantine and later p=reject. Your domain can have only one SPF record. For Google Workspace it looks like v=spf1 include:_spf.google.com ~all. You can test your setup with the email spoofing checker.
How to verify each fix
A fix is not done until you have proof. Do this for every finding:
- Back up first. Take a full site and database backup before any change.
- Re-run the exact steps from the report. Good reports include a URL, a request or a screenshot showing how the issue was reproduced. Repeat that, and the problem should no longer happen.
- Use a simple command where it fits. All of these are for sites you own:
- Headers:
curl -sI https://example.com | grep -i -E 'strict-transport|x-content-type|referrer-policy' - DMARC:
dig +short TXT _dmarc.example.com - Exposed file:
curl -s -o /dev/null -w '%{http_code}\n' https://example.com/.env(you want 403 or 404, not 200)
- Check the page content, not just the code. Some sites answer 200 for every URL with a friendly error page.
- Click through the normal flows. Log in, check out, submit a form. Security fixes sometimes break features.
- Save before and after evidence. A short note with the date, who fixed it and a screenshot or command output is enough.
Ask the tester for a retest
A retest is when the tester checks the same issues again and confirms they are closed. It is the difference between "we think it is fixed" and "someone independent confirmed it." Ask once your batch is done, not after every single change.
Send the tester:
- The report name and date.
- A list of finding IDs, each marked Fixed, Mitigated or Accepted, with one line on what you did.
- Any URLs or accounts that changed.
- Whether the site is live or on staging.
A short message works well:
> Hi, we have fixed findings 1, 2, 4 and 6 and mitigated finding 3. Finding 5 is accepted for now. Could you retest these and send an updated report? Is the retest included in the original price, and how long will it take?
Ask for an updated report that marks each finding as fixed or still open. Keep it. It is useful when a customer sends you a security questionnaire.
If you cannot fix something yet
Sometimes a finding needs a rewrite or a vendor update you do not control. Do not ignore it. Reduce the risk (block the path, restrict by IP, switch the feature off) and write down the decision: who accepted the risk, why, and a review date. A written, dated decision is far better than a silent gap.
Key takeaways
- Fix what outsiders can exploit without logging in first, especially anything near customer data or admin access.
- Use the severity table for deadlines, then adjust for your business and for known exploited CVEs.
- Group findings by root cause so one update can close several items.
- Verify every fix by repeating the original steps, and keep before and after evidence.
- Ask for a retest and an updated report once a batch is done.
Need a human to find the findings?
If you do not have a report yet, a person reading your site the way an attacker would is the honest next step. Our $49 Circuit audit is a manual review of a small site with a written report: severity, evidence and exact fixes for every issue. It is not an enterprise pentest or a compliance certificate. See pricing, or read how manual and automated testing differ first.
A real human security engineer audits your whole site by hand and sends a full report: every issue, its severity, and the exact fix. From $49, with a 14-day money-back guarantee.
See pricing