
A usable penetration test report is one your leadership can skim and your engineers can act on. It restates the scope you agreed, explains how testing was done, and lists each confirmed finding with a severity, evidence, business impact in plain language, and remediation a developer can follow. This post shows how to read that report using one annotated hypothetical finding, then how to triage fixes and verify them before a retest. A penetration test is not a compliance certification, and the report is part of how you judge whether the engagement was usable.
Prefer a worked example first? Jump to the clickable sample report.
Short answer
Read the executive summary for the picture, then read each finding as if you have to assign the fix. A usable finding names where it was observed, includes steps a developer can follow, shows evidence of what was confirmed, ties impact to a business asset, and recommends a concrete change. If those parts are missing, the readout is where you ask for them. After the readout, turn each confirmed finding into a ticket, verify the fix against the original steps, then use the retest terms in the statement of work when you are ready for the tester to re-check.
What a report usually contains
Layouts vary. These parts are the ones buyers actually use. If one is thin or missing, that is a signal, not a style preference.
- Executive summary. A short section for leadership: what was tested, how many confirmed findings sit at each severity, and which issues need a decision first. Check that it matches the technical findings. The counts and the top issues should not disagree.
- Technical findings. The body of the report. Each confirmed issue should stand on its own so a developer can pick it up without reading the rest of the document.
- Severity method. How ratings were assigned, for example a named scale such as Critical, High, Medium, and Low, and what those labels mean in this engagement. Context matters: the same class of bug can be High in one system and Medium in another. A critical-severity issue is one the tester treats as urgent, often by contacting the named person during testing rather than waiting for the report.
- Evidence. What the tester observed, such as a response, a screenshot description, or a log line, that shows the issue is real. Scanner output on its own is not enough.
- Business impact. What it means for your customers, data, or operations, written so someone who does not test systems for a living can follow it.
- Remediation. The change that would close the issue, specific enough to implement or to ticket.
- Methodology notes. Scope, testing window, authorization, what was in and out of scope, and any limits, for example no denial-of-service testing. Confirm this matches what you agreed before testing started. Authorized testing only happens inside that written scope.
Those parts are easier to judge if you read them in a stable order: scope and limits first, then the executive summary, then each finding, then an attack path if the report includes one, then the work of triage and retest. The annotated summary below shows what that second step can look like.
How to read the executive summary: an annotated illustrative example
Illustrative only, not a client or past engagement. Example Corp, app.example.com, and staff.example.com are placeholders.
Most people jump to the findings. Start with the executive summary, after you have checked that the scope matches what you agreed. A usable summary tells a leadership reader what was tested, what the tester confirmed in plain language, how many findings sit at each severity, and what to do first. If the counts or the lead issue disagree with the technical sections, that is a cue to ask in the readout, not a reason to treat either part as optional.
In this example, Example Corp asked for an authorized test of its customer application at app.example.com and the staff console at staff.example.com, in the agreed staging environment. The same engagement is used in the annotated finding and the triage queue later on this page. The summary below is hypothetical. It is here so you can see what to check.
Illustrative only, not a client or past engagement
A stable order for reading a report
- 1
Scope and limits
Confirm what was tested and what was left out.
- 2
Executive summary
Get the picture and the first decision, in plain language.
- 3
Each finding
Check evidence, impact, and a fix a developer can follow.
- 4
Attack path, when present
See whether findings matter more together than alone.
- 5
Triage and retest
Turn confirmed findings into tickets. Retest terms set in the statement of work.
Illustrative executive summary
Example Corp
Scope and window
Authorized test of the customer application at app.example.com and the staff console at staff.example.com, in the agreed staging environment. Testers used the logged-out view plus the supplied support and admin test accounts. Written authorization came from Example Corp, which controls those hosts. Denial-of-service testing stayed out of scope.
What to check: The hosts, environment, and roles should match the statement of work you agreed. An unmentioned system is untested, not cleared.
Plain-language result
Testing confirmed that a support impersonation link used to view a customer workspace kept working after the support session ended. Anyone who captured that link could keep acting as that customer on app.example.com. Two other confirmed findings were less urgent and did not depend on this one.
What to check: A leadership reader should be able to follow this without the technical sections. The lead issue should match a confirmed finding later in the report.
Confirmed findings by severity
0
Critical
1
High
1
Medium
1
Low
These are counts for this illustrative report, not a grade of Example Corp.
What to check: The counts should match the findings section. A disagreement is a cue to ask in the readout, not a reason to ignore either number.
What to do first
Invalidate the resume token when the support session ends, then verify the original steps no longer open a customer workspace before asking for a retest. Retest terms set in the statement of work.
What to check: The first action should point at a confirmed finding, not a generic improvement, and it should be specific enough to ticket.
Limits and what was not covered
Production data, hosts not named in the statement of work, and denial-of-service testing were out of scope. This summary does not include an attack path, because the three confirmed findings do not rely on each other.
What to check: A quiet result in an area that was not in scope does not mean that area was tested. If an attack path is present in your report, read it next to the individual findings.
Once the summary is consistent with the rest of the report, read each finding as if you have to assign the fix. The next section annotates the High issue named above. This illustrative summary does not include an attack path, because the three confirmed findings do not rely on each other. If your report does include one, read it after the individual findings so you can see whether two issues matter more together than they do alone.
How to read one finding: an annotated illustrative example
This is an illustrative scenario, not a real client or case study. Example Corp, staff.example.com, and app.example.com are placeholders.
Service pages on this site show how an engagement runs from scope to readout. This section is different: it is one finding, annotated so you can see what to look for when you open a report. The issue is hypothetical. A real report covers every confirmed finding in the same shape.
In this example, Example Corp asked for an authorized test of its customer application at app.example.com and the staff console at staff.example.com. Support users can view a customer workspace to help with tickets. The tester confirmed that the link used for that view kept working after the support session ended.
Illustrative only, not a client or past engagement
Illustrative finding card
Example Corp
Title
Support impersonation link stays valid after the session ends
Severity and how it was rated
High. Rated High because the resume link reuses a support session against a customer workspace.
Affected asset
The support console at staff.example.com, action labeled View as customer, then the resume URL at https://app.example.com/support/resume.
Description
The tester confirms that the link used for that view keeps working after the support session ends.
Evidence and reproduction steps
Using the agreed support-role test account, sign in at https://staff.example.com, open the agreed test customer workspace, choose View as customer, copy the resume URL, sign out, and open that URL in a fresh browser. The resume URL still issues a session cookie after the support console shows the impersonation as ended, and the session opens as customer-test@example.com on app.example.com.
Impact
Anyone who captured that link, from a screenshot, a shared screen, a proxy log, or a referrer, could keep acting as that customer after support work finished, including reading or changing that customer's records.
Remediation
Invalidate the token when the support session ends, and make it single-use and short-lived. Bind it to the support user's session rather than putting a reusable secret in the query string.
Retest status
Not yet retested. Retest terms set in the statement of work.
Illustrative finding (hypothetical)
Support impersonation link stays valid after the session ends
Title
Support impersonation link stays valid after the session ends
What to look for: A specific issue on a named host, not a generic heading such as Access Control with no target.
Severity
High
What to look for: A label from a stated scale (Critical, High, Medium, or Low), plus a short reason that fits this system, not a score with no explanation.
Where observed
The support console at staff.example.com, action labeled View as customer, then the resume URL at https://app.example.com/support/resume.
What to look for: A hostname, path, and role a developer can find, using the in-scope names, not only "the application."
Steps to reproduce
- Using the agreed support-role test account, sign in at
https://staff.example.com. - Open the agreed test customer workspace for Example Corp and choose View as customer.
- Copy the resume URL from the address bar, which looks like
https://app.example.com/support/resume?token=EXAMPLETOKEN. - Sign out of the support console and wait until that impersonation is marked ended.
- In a fresh browser with no
staff.example.comcookies, open the copied URL. - The session opens as
customer-test@example.comonapp.example.com, with access to that workspace's records.
What to look for: Numbered steps a developer can follow in the agreed test accounts, without needing the tester on the call.
Evidence summary
The resume URL still issued a session cookie after the support console showed the impersonation as ended. The workspace that loaded belonged to the agreed test customer, not the support user. Illustrative request and response:
GET /support/resume?token=EXAMPLETOKEN HTTP/1.1
Host: app.example.com
HTTP/1.1 302 Found
Location: /workspace/example-corp
Set-Cookie: session=...What to look for: What was actually observed, such as a response, a token that still worked, or a record that loaded, not only a scanner plugin name.
Business impact
Anyone who captured that link, from a screenshot, a shared screen, a proxy log, or a referrer, could keep acting as that customer after support work finished, including reading or changing that customer's records.
What to look for: A plain-language effect on a business asset, such as a customer account or workspace, not only a technical identifier.
Remediation guidance
Invalidate the token when the support session ends, and make it single-use and short-lived. Bind it to the support user's session rather than putting a reusable secret in the query string. Log every impersonation start and end. Require a fresh step-up on staff.example.com before starting a new impersonation.
What to look for: An actionable change a developer can ticket, not a generic reminder to follow best practices.
A weak write-up of the same issue might say only “Broken authentication, High, improve session management.” That does not tell a developer where to look, how to confirm it, or what to change. If a finding in your report is a title and a severity with nothing under them, ask for the missing parts in the readout rather than guessing.
Severity labels only make sense next to the method the report names. This tiny table is illustrative for the same hypothetical report. It is not a universal scale.
| Label | What it meant in this example |
|---|---|
| Critical | Confirmed, easy-to-use access to another customer's live data. In this kind of report, a critical-severity issue is raised with the named contact during testing rather than waiting for delivery. |
| High | Confirmed access that should be treated as urgent once the report lands. The impersonation link in this example is rated High because it reused a support session against a customer workspace. |
| Medium | A confirmed issue that needed extra conditions, a narrower role, or more steps, so it is less urgent than High in this example. |
When you read your own report, check that the rating on each finding matches the impact paragraph beneath it. A High title with Low impact text, or the reverse, is a cue to ask about the method in the readout.
Walk through an illustrative sample report
Illustrative only, not a client, case study, or past engagement. Example Corp, app.example.com, and staff.example.com are placeholders.
The annotated summary and finding above use the same hypothetical Example Corp engagement. This sample puts those parts in one place, and adds a conclusion and a retest note. Click each section to see a usable write-up, then what is missing when that part is thin. Use the checklist on a report you already have. This is a reading aid, not a grade of Example Corp, and not a substitute for reading your own report.
In this example, the three confirmed findings do not rely on each other, so the sample does not invent an attack path. If findings in your report do chain, the report should show that path with hosts, evidence, and a first change that would break the chain.
Illustrative only, not a client, case study, or past engagement.
Click through a sample report
Each section shows a usable write-up for the illustrative Example Corp engagement. Switch to the thin version to see what is missing. Example Corp, app.example.com, and staff.example.com are placeholders.
A stable order for reading this sample
Illustrative sample report
Example Corp
Executive summary
Scope and window
Authorized test of the customer application at app.example.com and the staff console at staff.example.com, in the agreed staging environment. Testers used the logged-out view plus the supplied support and admin test accounts. Written authorization came from Example Corp, which controls those hosts. Denial-of-service testing stayed out of scope.
Plain-language result
Testing confirmed that a support impersonation link used to view a customer workspace kept working after the support session ended. Anyone who captured that link could keep acting as that customer on app.example.com. Two other confirmed findings were less urgent and did not depend on this one.
Confirmed findings by severity
Critical 0, High 1, Medium 1, Low 1. These are counts for this illustrative report, not a grade of Example Corp.
What to do first
Invalidate the resume token when the support session ends, then verify the original steps no longer open a customer workspace before asking for a retest. Retest terms set in the statement of work.
What good looks like
A leadership reader can see what was tested, the lead confirmed issue in plain language, how many findings sit at each severity, and a first action specific enough to ticket. The counts and the lead issue should match the findings section.
What is missing when this is thin
No named hosts or environment, no counts by severity, no lead issue a reader can match to a finding, and no first action specific enough to ticket.
Completeness checklist
Use these questions on a report you already have, not as a score of the sample above. Mark each item as you read.
0 of 5 items marked.
A clickable sample report and checklist by The PenTesting Company. Example Corp, app.example.com, and staff.example.com are placeholders. In this illustrative report, the three confirmed findings do not rely on each other. A usable report says that, rather than inventing a chain. If findings in your report do chain, the report should include an attack path that names hosts, evidence, and a first change that would break the chain. Not every usable report contains a chain. Retest terms set in the statement of work. A penetration test is not a compliance certification.
What to ask in the readout
The readout is the call after the report is delivered. Bring the people who will fix the issues, not only the people who bought the test. Use the time to resolve anything a developer would get stuck on.
- Walk us through this finding's evidence again, in the same environment we will use to fix it.
- What would you fix first, and why that order rather than severity order alone?
- Which agreed systems were not fully covered, and what stayed out of scope?
- If we make the change as written, what nearby behavior should we re-check?
- Who on your side answers questions while we implement the fix?
- For any critical-severity issue, when were we told during testing, and what had already been done by the time the report arrived?
You do not need to replay every step on the call. You do need enough shared understanding that a ticket can be written without guessing.
After the readout: triage and verify
Reading the report and asking good questions on the readout are the first half. The work that usually decides whether the engagement was useful is what your team does next: turn findings into tickets, pick an order that matches business risk, verify each fix against the original steps, and only then ask for a retest under the statement of work.
- One ticket per confirmed finding. Copy the host or path, the steps to reproduce, the suggested remediation, and a short impact line into the ticket so a developer does not need the full PDF open.
- Order by business risk, not labels alone. Severity is a starting point. A High that reaches customer data usually comes before a High that only affects an unused admin screen. Ask in the readout which issues the tester would fix first and why.
- Verify before you request a retest. Have someone other than the author of the change replay the report steps in the agreed environment. If the original path still works, the finding is not ready for the tester to re-check.
- Keep Low findings visible. They often share a root cause with higher ones, or become easy wins in the same release. Do not hide them because the queue is long.
Worked example: Example Corp triages the report
Illustrative only, not a client or past engagement. Example Corp, staff.example.com, app.example.com, and the findings below are placeholders. This shows how one hypothetical team might turn a report into tickets and verification steps after the readout.
Say Example Corp has just finished the readout for an authorized test of app.example.com and staff.example.com. The engineering lead and the product owner sit with the report open. They would open one ticket per confirmed finding, write the verification steps into the ticket, and agree who owns each fix before anyone asks for a retest.
| Finding | Severity | Ticket focus | Verify before retest |
|---|---|---|---|
| Support impersonation link stays valid after the session ends | High | Invalidate the resume token when support ends, bind it to the support session, and add a regression test that fails if the old URL still opens a customer workspace. | Replay the report steps with the agreed support test account after the change. Confirm the ended session cannot resume, and that a fresh impersonation still works for support. |
| Display name rendered unsafely in the admin user list | Medium | Encode the display name on output in the admin user list, and add a unit test with a script-shaped name. | Create a test user whose display name includes a simple script tag, open the admin list as the agreed admin account, and confirm the page shows the text without executing script. |
| Password reset email still names the product build id | Low | Remove the build id from the reset email template. Keep the change with the next template deploy. | Trigger a reset for a test inbox and confirm the message no longer includes the build id. |
In this example, Example Corp would ship the impersonation-token fix first because it reaches customer workspaces, then the admin list encoding, then the email template cleanup. Before anyone requests a retest, each owner would replay the report steps on the build that includes their change. Retest terms set in the statement of work. A penetration test is not a compliance certification. Closing these tickets does not make Example Corp compliant on its own.
Retest and what “fixed” means
Retest terms set in the statement of work.
A retest usually means the tester returns to the original confirmed findings, in the agreed scope, and tries the same steps again. “Fixed” in that sense means the original issue could not be reproduced as written. It does not mean every related path is gone, and it does not mean the rest of the system is free of other issues. Anything found after the testing window, or outside the agreed scope, is a different conversation.
Ask for that retest when your team has already verified the original steps in the environment the tester will use, and you can name the builds or hosts that include the fixes. If you still need to confirm which release holds a change, wait. A retest that lands on the wrong build wastes the window the statement of work set aside for it.
A penetration test is not a compliance certification. The report can be one input to a vendor review or a broader security program. It does not certify that you meet a standard.
If you want to see how we write findings for a scoped engagement, read about web application security testing or network penetration testing. Contact us to agree a scope.
Frequently asked questions
How do I know if a report is complete enough to hand to developers?
If each confirmed finding names where it was observed, includes steps to reproduce, shows evidence, states business impact, and gives a concrete fix, a developer can usually start. If any of those parts are missing, ask for them in the readout before you file tickets.
Does a report with only Low findings mean we are secure?
No. A penetration test is a point-in-time look at the systems and window you agreed. It does not mean the system is free of issues, and it does not speak to changes made after testing.
Is a penetration test report a compliance certification?
No. A penetration test is not a compliance certification. The report can support a compliance program or a vendor review by providing independent evidence of testing. It does not make you compliant on its own.
What should I do if a severity rating feels off?
Ask how the rating was assigned and which part of the impact drove it. Align on whether the business asset at risk matches the label. It is reasonable to discuss context. It is not useful to argue a label down without changing the facts.
What does a retest actually cover?
Retest terms set in the statement of work. In most engagements it is a return to the original confirmed findings, not a second full test of everything in scope.
How should we turn findings into tickets after the readout?
Give each confirmed finding its own ticket with the host or path, the steps to reproduce, the suggested fix, and a short note on business impact. Order the queue by what would hurt customers or operations first, not by severity labels alone. Keep Low items visible so they do not disappear, and assign an owner who can say when a fix is ready to verify.
When should we ask for a retest after we fix findings?
Retest terms set in the statement of work. Ask once your team has verified the original steps no longer work in the agreed environment, and you can point the tester at the builds or hosts that include those fixes. A retest usually re-checks the original confirmed findings. It is not a second full test of everything in scope, and it does not mean the rest of the system is free of other issues.

Have a system you want tested?
Tell us what you are working with and we will follow up with next steps.