Skip to main content
Sample reportReviewed

Penetration test report example: a redacted lab report, annotated section by section

A finished web application penetration test report from a lab application Clever Ops controls, with credentials and private values redacted.

Each part shows what the report says, what a good report should show there, and what to ask a provider whose report leaves it out. It is a lab example, not client work and not independent assurance.

About this lab report, and why it is labelled

A pen test report example is only useful if you can see the real findings, the real limits and the real status, not a brochure version of them.

This page walks through a complete report Clever Ops wrote about crm.lab.test, a CRM application in a Clever Ops lab. The application holds two seeded test tenants, Northstar and Cedar, filled with test data. Nothing in it belongs to a client or to any real business, so the report is evidence about a lab and nothing more.

Because Clever Ops controls the application, it is not independent assurance either. What it shows is the format of a report, and how a report handles its own limits.

The report is identified as CO-2026-SEC-001, version 1.0, with a final status. Its overall status line reads: five rated findings open, one informational observation. None of the findings has been retested. Every identifier, title, rating, score and status below is shown as the report states it, including the parts a buyer might wish were stronger, because a sample that hides its own gaps teaches the wrong lesson.

Redaction follows the report's own rule. Session cookies, the exposed service credential, memo text and contact details appear here as (redacted). Object identifiers such as ct_cedar_201 are kept, because a developer needs them to trace a request. Request and response excerpts keep their shape, so you can see what a reproducible finding looks like without seeing a secret.

One example cannot show every part a report can contain. This one has no change history beyond its first version, no checklist mapping the work to a published testing guide, no build identifier for the application and no retest results.

Those gaps are named where they occur, and the yardstick table on this page sets out what to look for in each part of any report, including reports from other providers.

The executive summary and findings at a glance

The summary is written for the person who decides what gets fixed first.

In the lab report it opens with the two problems that matter most: customer records were exposed across tenant boundaries, and a member could assign themselves an administrator role. It then counts the rest. Beside the Critical finding sit two High and two Medium findings, plus one observation recorded without a risk rating because further testing narrowed its impact.

It names the priority in a phrase, systemic server-side authorisation, and explains it: object ownership and role changes need to be enforced by the API on every relevant request, rather than inferred from identifiers the client sends or from how the interface behaves. It also puts the verification limit on the stored cross-site scripting (XSS) finding in the executive summary instead of burying it.

The stored HTML and the raw rendering path were found, but script execution was not visually confirmed, because a browser was not available during the assessment.

The summary ends with a business response in five steps: fix tenant authorisation and role assignment before treating the application as ready for further exposure; rotate the exposed service credential and take credential handling out of the public browser bundle; replace or sanitise raw note HTML, then check its behaviour in a supported browser; add two-tenant and two-role regression tests to the development workflow; and arrange one consolidated retest once the fixes are deployed to the agreed test build.

Findings at a glance, as the lab report lists them

Each line gives the identifier, title, business rating, CVSS 3.1 score and status exactly as the report's summary table shows them. The summary table prints the CVSS band beside the score only for WS-01, the one finding whose two ratings differ.

  • WS-01, Cross-tenant contact read and modification. Business rating: Critical. CVSS 3.1: 8.1 High. Status: open, not retested.
  • WS-02, Member can self-assign an administrator role. Business rating: High. CVSS 3.1: 8.1. Status: open, not retested.
  • WS-03, Stored XSS in note body. Business rating: High. CVSS 3.1: 8.7. Status: open, browser confirmation outstanding.
  • WS-04, Service credential exposed in public JavaScript. Business rating: Medium. CVSS 3.1: 5.3. Status: open, not retested.
  • WS-05, CSV formula injection in contact export. Business rating: Medium. CVSS 3.1: 4.6. Status: open, not retested.
  • O-01, Foreign-tenant contact reference accepted for notes. Rating: Info. CVSS 3.1: not scored. Status: open, impact narrowed by follow-up testing.

What to check in any executive summary

Read the summary on its own and ask whether it would let a director decide what to fund first. The OWASP Web Security Testing Guide (WSTG) v4.2 offers its reporting chapter as suggestions rather than strict rules, and it likens the executive summary to the elevator pitch of the report: the objective of the test, the business need behind it, key findings in a business context, and strategic recommendations, without jargon.

A summary that lists vulnerability names with no consequence, or leaves a known limit for the back pages, falls short of that.

The lab summary covers the findings, the priority and the response well. It does not state the business need behind the test, which a lab application does not have. A report on your own system should say why the test was commissioned, because that reason decides which findings matter most.

Scope and limitations: the boundary every finding sits inside

A finding is only as broad as the scope it was found in.

The lab report names the application, crm.lab.test, as an HTTPS web application with a JSON API, and the two seeded tenants. Its role list has one starting role, a Northstar member, and notes administrator behaviour observed after the privilege escalation in WS-02. The functional areas are contacts, notes, exports, the profile, the administrator overview and the public JavaScript assets.

The perspective is signed-in testing plus a review of the browser code the application serves publicly, and the stated focus is tenant isolation, object-level authorisation, role boundaries, stored content and export handling.

Seven items are listed as out of scope or not performed:

  • Denial-of-service, load and volumetric testing.
  • Destructive deletion of tenant records. DELETE /api/contacts/{id} was not exercised.
  • Third-party services, the underlying network infrastructure and host operating systems.
  • Social engineering, physical testing and incident response.
  • Source-code review beyond the browser assets the application serves publicly.
  • Compliance certification, or any statement that the application meets a regulatory standard.
  • A complete inventory and test of every application route. Route coverage was limited to the functional areas listed.

What a scope section should let you answer

Three questions: what was tested, as whom, and what was left out on purpose. WSTG v4.2 suggests a report introduction that sets out the scope agreed with the organisation and its limitations, which it says can be out-of-bounds areas, broken functionality, lack of cooperation, lack of time, and lack of access or credentials.

The lab report answers the first and third questions clearly. Its exclusions even rule out the reading that every route was covered. It is thinner on the second: the role list names one starting role and no Cedar account, so every cross-tenant result is seen from the Northstar side.

It also does not identify the build or release of the application that was tested. A report on your system should name the build, because a finding, a negative result and a retest each describe one build, and the comparison only works if you know which.

Reproducible evidence: WS-01 from request to restore

Evidence is what turns a finding from an opinion into something your developer can repeat.

WS-01 shows the shape. Its description says the contact API accepted a Cedar-owned contact identifier while the request was authenticated as a Northstar member, returned Cedar contact data including a confidential memo field, and then accepted a PATCH request that modified the same Cedar record. The excerpts below keep the report's structure, with the secrets replaced.

  • Read request: GET /api/contacts/ct_cedar_201 HTTP/1.1, Host: crm.lab.test, Cookie: crm_session=(redacted), sent from a Northstar member session.
  • Read response: 200 OK, returning a Cedar-owned contact object with id ct_cedar_201 and tenant_id cedar, its name and private_memo values redacted.
  • Write request: PATCH /api/contacts/ct_cedar_201 in the same session, Content-Type: application/json, with a body setting stage to Proposal and private_memo to a test marker.
  • Write response: 200 OK, reflecting the values the attacker controlled. The original values were restored after testing, and cross-tenant deletion was not attempted.
  • Contrasting check: POST /api/contacts forced the session tenant and did not honour a tenant override, which the report reads as inconsistent enforcement rather than an absence of tenant handling throughout the application.

The sentence that stops a symptom fix

WS-01 also separates the cause from what makes the cause easier to exploit. The identifiers seen in the export follow a predictable pattern in some cases, and predictable identifiers would make enumeration easier. The report then says plainly that random identifiers would not address the missing authorisation check.

Without that sentence, a developer under time pressure could swap the identifier format, watch the reproduction fail by chance, and close a finding that is still open.

How the lab report says its evidence was recorded

Above the findings, the report states its evidence provenance: the request and response examples are reconstructed from contemporaneous tester notes and redacted, and they are not presented as matched byte-for-byte packet captures. Its evidence-handling section adds that matching raw reproduction captures were not available for every finding and are not claimed.

That is a genuine limitation, stated where a reader will meet it. The excerpts show what was sent and what came back, as noted at the time, but they are not the raw traffic.

WSTG v4.2 suggests that the findings section give the technical team all the necessary information to understand the vulnerability, replicate it and resolve it, and that possibly sensitive data such as passwords, personal information or credit card details be masked. When you read any provider's report, ask whether the excerpts were captured during the test or written up afterwards, and whether raw captures will be kept for the retest.

Then pick one finding and check that you can see the account, the request, the response and the contrasting request that shows the boundary was crossed.

Severity rationale: one score, two ratings, and the reasons between them

The lab report rates each finding twice.

CVSS 3.1 measures technical severity. A business-risk rating weighs the data held, how exposed the affected function is, the controls already in place and the cost of interruption. The report's severity bands match the CVSS v3.1 qualitative scale (Low 0.1 to 3.9, Medium 4.0 to 6.9, High 7.0 to 8.9, Critical 9.0 to 10.0) and add Informational for observations that are not scored.

Every score comes with its vector string. The FIRST specification describes the vector string as a text representation of a set of CVSS metrics that begins with the label CVSS: and the version number. The vector is what makes a score checkable: anyone can see the inputs and dispute one. These are the five vectors in the lab report.

  • WS-01, 8.1:CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:N
  • WS-02, 8.1:CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:N
  • WS-03, 8.7:CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:C/C:H/I:H/A:N
  • WS-04, 5.3:CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:N
  • WS-05, 4.6:CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:U/C:L/I:L/A:N

Reading the WS-01 vector

Each pair is a metric and its value, as defined in the FIRST CVSS v3.1 specification. 1 min read

Same vector, different business rating: WS-01 and WS-02

WS-01 and WS-02 share one vector and one score, 8.1, yet the report rates WS-01 Critical and WS-02 High for the business. 1 min read

How the other scores carry their limits

The lower findings show something just as useful: a score that stays inside what was demonstrated. 1 min read

What a good severity section gives you

Look for the scale defined, a vector beside every score, and a written reason wherever the business rating departs from the technical one. 1 min read

Category mappings: ask for the basis

The lab report gives each finding a CWE identifier and maps four of the five rated findings to the OWASP Top 10:2025: WS-01 and WS-02 to A01 Broken Access Control, and WS-03 and WS-05 to A05 Injection. 1 min read

Report yardstick: twelve parts of a pentest report to check

One row for each part of a penetration test report. Read across a row for what a good report shows there, the question to put to a provider whose report lacks it, what the Clever Ops lab report shows and what it does not.

How to use it

To compare providers, keep the first three columns and fill in the last two for each sample you are sent. The lab report covers crm.lab.test, a lab application Clever Ops controls; it is not client work, not independent assurance and not evidence about any real business.

12 rows

Report yardstick: twelve parts of a pentest report to check

Part of the reportWhat a good report showsAsk the provider if it is missingWhat the lab report showsWhat the lab report does not show
Document control and labellingA report ID, version and status, the target, the type of assessment and the environment, with dates and a change history so every later copy can be matched to the one you hold.Which version is this, what changed since the last one, and was the target production, staging or a lab?Report ID CO-2026-SEC-001, version 1.0, final status, the target crm.lab.test, the assessment type, a Clever Ops lab environment, and an overall status line of five rated findings open and one informational observation.A change history, since this is the first version, and any build or release identifier for the application tested.
Executive summaryThe objective and business need, the key findings and their consequence in plain words, and what to do first, readable without the rest of the report.Could someone who never reads the findings decide what to fix and fund from this page alone?The two most serious problems in one sentence, the count by rating, the priority of server-side authorisation, the WS-03 verification limit and a five-step business response.The business need behind the test, which a lab application does not have.
Scope and exclusionsThe targets, tenants or customer organisations, roles and accounts, functional areas, testing perspective, and a list of what was out of scope or not performed.Which roles and accounts were used, which features were covered, and what was deliberately left out?One application and its JSON API, two seeded tenants, the Northstar member role, six functional areas, and seven exclusions, including deletion of tenant records and a complete inventory of every route.A Cedar account or a second starting role, so cross-tenant results are seen from the Northstar side only.
Limitations and verification limitsEvery place the evidence stops short of proof, stated beside the finding it affects and repeated in the summary when it changes a rating.Is any impact inferred rather than observed, and does the report say so where that finding appears?A browser was not available, so WS-03 script execution was not visually confirmed; cross-tenant deletion was not attempted; evidence excerpts are reconstructed from contemporaneous tester notes.Limits caused by time or by missing cooperation, two other kinds the OWASP reporting guidance lists.
Findings summaryOne line per finding with its reference ID, title, rating and status, so the whole result fits on one screen.Do the same IDs carry through the details, the remediation plan and the retest?Six lines, WS-01 to WS-05 and O-01, each with a business rating, a CVSS 3.1 score or "not scored", and a status.Any status after a retest, because nothing has been retested.
Evidence for each findingThe request and response, the account and tenant that sent them, a contrasting request where one helps, and secrets masked, enough for your developer to replicate the result.Were these excerpts captured during the test or written up afterwards, and will raw captures be kept for the retest?Redacted request, response or export excerpts for every rated finding, object IDs kept for tracing, a contrasting POST /api/contacts check for WS-01, and a note that changed values were restored.Byte-for-byte captures, screenshots or per-request timestamps. The report says matching raw captures are not claimed.
Severity rating and scoringThe scale defined, a CVSS vector string beside every score, and a business rating with a written reason wherever it departs from the technical band.What inputs produced this score, and why is the business rating higher or lower than the technical one?Severity bands, five CVSS 3.1 Base scores each with its vector, a business rating on every rated finding, and the reason WS-01 is Critical while scoring 8.1 High.CVSS Temporal or Environmental scores, which the FIRST specification says consumers should add for their own environment.
Weakness class and categoryA CWE and an OWASP Top 10 category on every finding, with the basis for any mapping that departs from OWASP's own list.Which category does each finding fall under, and how was it chosen?A CWE on every finding; A01 Broken Access Control for WS-01 and WS-02, and A05 Injection for WS-03 and WS-05.A category for WS-04 or O-01, or a reason for placing WS-02 under A01 when OWASP lists its CWE, CWE-915, under A08.
Remediation guidance and planFixes aimed at the cause and written for the stack, a plan in priority order, and tests that keep the problem from returning.Does each fix close the cause, or only the request that was shown?Numbered fixes for every rated finding, a four-band plan by assessed risk, and the warning that random identifiers would not repair WS-01.Effort or delivery dates. The plan leaves timing to exposure, data sensitivity, release dependencies and engineering capacity.
Negative resultsChecks that found nothing, each tied to the request, account and build it covers, with untested areas listed separately.Which areas came back clean, on which build, and with which accounts?Eight checks with observed results, including tenant overrides that were not honoured and an admin settings route that returned 404, plus the caveat that each covers only what was tested.A build identifier for those results to refer to.
Test data and clean-upWhat was changed, created or left behind, what was restored, and what the owner still needs to remove.Is anything from the test still in the system, and how do we find it?Restored Cedar contact values and member role, a deleted CSV probe contact, and two XSS test notes left in place because the application had no note-deletion route.Identifiers for the two leftover notes, which an owner would need to find them.
Retest criteria and statusStatus words defined before any retest, criteria for each finding written in advance, and an updated report that keeps findings not retested apart from fixed ones.What will the retest check for each finding, and how will a finding we chose not to fix be shown?Every finding open at issue, four defined statuses, the rule that an item not retested is not presented as passed, criteria written out for WS-01, and a browser confirmation required for WS-03.A completed retest or any change of status, and separate retest criteria for WS-02, WS-04, WS-05 and O-01.

Part of the report

What a good report shows

  1. Document control and labelling
    What a good report shows
    A report ID, version and status, the target, the type of assessment and the environment, with dates and a change history so every later copy can be matched to the one you hold.
    Ask the provider if it is missing
    Which version is this, what changed since the last one, and was the target production, staging or a lab?
    What the lab report shows
    Report ID CO-2026-SEC-001, version 1.0, final status, the target crm.lab.test, the assessment type, a Clever Ops lab environment, and an overall status line of five rated findings open and one informational observation.
    What the lab report does not show
    A change history, since this is the first version, and any build or release identifier for the application tested.
  2. Executive summary
    What a good report shows
    The objective and business need, the key findings and their consequence in plain words, and what to do first, readable without the rest of the report.
    Ask the provider if it is missing
    Could someone who never reads the findings decide what to fix and fund from this page alone?
    What the lab report shows
    The two most serious problems in one sentence, the count by rating, the priority of server-side authorisation, the WS-03 verification limit and a five-step business response.
    What the lab report does not show
    The business need behind the test, which a lab application does not have.
  3. Scope and exclusions
    What a good report shows
    The targets, tenants or customer organisations, roles and accounts, functional areas, testing perspective, and a list of what was out of scope or not performed.
    Ask the provider if it is missing
    Which roles and accounts were used, which features were covered, and what was deliberately left out?
    What the lab report shows
    One application and its JSON API, two seeded tenants, the Northstar member role, six functional areas, and seven exclusions, including deletion of tenant records and a complete inventory of every route.
    What the lab report does not show
    A Cedar account or a second starting role, so cross-tenant results are seen from the Northstar side only.
  4. Limitations and verification limits
    What a good report shows
    Every place the evidence stops short of proof, stated beside the finding it affects and repeated in the summary when it changes a rating.
    Ask the provider if it is missing
    Is any impact inferred rather than observed, and does the report say so where that finding appears?
    What the lab report shows
    A browser was not available, so WS-03 script execution was not visually confirmed; cross-tenant deletion was not attempted; evidence excerpts are reconstructed from contemporaneous tester notes.
    What the lab report does not show
    Limits caused by time or by missing cooperation, two other kinds the OWASP reporting guidance lists.
  5. Findings summary
    What a good report shows
    One line per finding with its reference ID, title, rating and status, so the whole result fits on one screen.
    Ask the provider if it is missing
    Do the same IDs carry through the details, the remediation plan and the retest?
    What the lab report shows
    Six lines, WS-01 to WS-05 and O-01, each with a business rating, a CVSS 3.1 score or "not scored", and a status.
    What the lab report does not show
    Any status after a retest, because nothing has been retested.
  6. Evidence for each finding
    What a good report shows
    The request and response, the account and tenant that sent them, a contrasting request where one helps, and secrets masked, enough for your developer to replicate the result.
    Ask the provider if it is missing
    Were these excerpts captured during the test or written up afterwards, and will raw captures be kept for the retest?
    What the lab report shows
    Redacted request, response or export excerpts for every rated finding, object IDs kept for tracing, a contrasting POST /api/contacts check for WS-01, and a note that changed values were restored.
    What the lab report does not show
    Byte-for-byte captures, screenshots or per-request timestamps. The report says matching raw captures are not claimed.
  7. Severity rating and scoring
    What a good report shows
    The scale defined, a CVSS vector string beside every score, and a business rating with a written reason wherever it departs from the technical band.
    Ask the provider if it is missing
    What inputs produced this score, and why is the business rating higher or lower than the technical one?
    What the lab report shows
    Severity bands, five CVSS 3.1 Base scores each with its vector, a business rating on every rated finding, and the reason WS-01 is Critical while scoring 8.1 High.
    What the lab report does not show
    CVSS Temporal or Environmental scores, which the FIRST specification says consumers should add for their own environment.
  8. Weakness class and category
    What a good report shows
    A CWE and an OWASP Top 10 category on every finding, with the basis for any mapping that departs from OWASP's own list.
    Ask the provider if it is missing
    Which category does each finding fall under, and how was it chosen?
    What the lab report shows
    A CWE on every finding; A01 Broken Access Control for WS-01 and WS-02, and A05 Injection for WS-03 and WS-05.
    What the lab report does not show
    A category for WS-04 or O-01, or a reason for placing WS-02 under A01 when OWASP lists its CWE, CWE-915, under A08.
  9. Remediation guidance and plan
    What a good report shows
    Fixes aimed at the cause and written for the stack, a plan in priority order, and tests that keep the problem from returning.
    Ask the provider if it is missing
    Does each fix close the cause, or only the request that was shown?
    What the lab report shows
    Numbered fixes for every rated finding, a four-band plan by assessed risk, and the warning that random identifiers would not repair WS-01.
    What the lab report does not show
    Effort or delivery dates. The plan leaves timing to exposure, data sensitivity, release dependencies and engineering capacity.
  10. Negative results
    What a good report shows
    Checks that found nothing, each tied to the request, account and build it covers, with untested areas listed separately.
    Ask the provider if it is missing
    Which areas came back clean, on which build, and with which accounts?
    What the lab report shows
    Eight checks with observed results, including tenant overrides that were not honoured and an admin settings route that returned 404, plus the caveat that each covers only what was tested.
    What the lab report does not show
    A build identifier for those results to refer to.
  11. Test data and clean-up
    What a good report shows
    What was changed, created or left behind, what was restored, and what the owner still needs to remove.
    Ask the provider if it is missing
    Is anything from the test still in the system, and how do we find it?
    What the lab report shows
    Restored Cedar contact values and member role, a deleted CSV probe contact, and two XSS test notes left in place because the application had no note-deletion route.
    What the lab report does not show
    Identifiers for the two leftover notes, which an owner would need to find them.
  12. Retest criteria and status
    What a good report shows
    Status words defined before any retest, criteria for each finding written in advance, and an updated report that keeps findings not retested apart from fixed ones.
    Ask the provider if it is missing
    What will the retest check for each finding, and how will a finding we chose not to fix be shown?
    What the lab report shows
    Every finding open at issue, four defined statuses, the rule that an item not retested is not presented as passed, criteria written out for WS-01, and a browser confirmation required for WS-03.
    What the lab report does not show
    A completed retest or any change of status, and separate retest criteria for WS-02, WS-04, WS-05 and O-01.

Remediation guidance and the order of work

Every rated finding in the lab report includes numbered fixes aimed at the cause.

For WS-01 they read: enforce tenant ownership in the data query or a shared authorisation layer, for example by requiring both the contact ID and the session-derived tenant ID to match; apply the same check to read, update, delete, export and indirect-reference operations; derive tenant context from the authenticated session and treat client-supplied tenant and object identifiers as untrusted; return a consistent 403 or 404 for foreign-tenant objects, according to the application's information-disclosure policy; add automated tests with at least two tenants and two roles for every object method; and review access logs for historical cross-tenant requests if the pattern has existed in a production system.

The other findings follow the same pattern. WS-02 moves role administration to a separate endpoint that checks the caller's authority and records an audit event. WS-03 renders note bodies as text unless rich HTML is a firm product requirement, and treats a Content Security Policy as defence in depth, not a substitute for removing the injection sink.

WS-04 starts with revoking or rotating the exposed credential and keeping service credentials on the server. WS-05 neutralises cells whose first non-whitespace character is =, +, - or @ in every user-controlled column, and asks whether confidential memo fields belong in the default export. O-01 validates that a referenced contact belongs to the caller's tenant before the note is inserted.

The report then groups the work into four bands by assessed risk, with no dates attached:

  • First priority: centralise tenant authorisation, restrict profile fields, move role changes to an authorised workflow, and rotate and remove the browser credential. WS-01, WS-02 and WS-04.
  • Next development cycle: remove or sanitise raw note HTML, neutralise spreadsheet formulas, validate note contact references and remove test artefacts. WS-03, WS-05 and O-01.
  • Before retest: deploy a stable test build, run two-tenant and two-role regression tests, confirm stored XSS behaviour in a browser and prepare evidence of credential rotation. All findings.
  • Ongoing: authorisation tests for new object routes, secret scanning for new bundles and export tests for new user-controlled columns. Prevention.

Judging the fixes in any report

The plan adds that delivery timing should account for exposure, data sensitivity, release dependencies and available engineering capacity, which leaves the schedule with the people who own the code. WSTG v4.2 suggests detailed steps on how to remediate each vulnerability, possible improvements that could help strengthen the security posture, and missing security practices.

When you read a provider's fixes, check that each one closes the cause rather than the single request that was shown, and that it comes with a test your developers can keep running after the report is filed.

Retest status: open, and saying so

No finding in the lab report had been retested at the time of issue, and the report says so.

The report defines the statuses a finding can take after a retest before any retest happens. Fixed: the original reproduction fails and the agreed regression checks pass. Partially fixed: some affected paths or impact remain reproducible. Still reproducible: the original issue remains exploitable in the agreed test conditions.

Not retested: the item was not included in the retest, or the access or build it needed was unavailable. It adds the rule worth looking for in any report: an item that was not retested is not presented as passed.

A consolidated retest covers the reported findings on an identified remediated build of the same application, within the agreed scope, and unrelated changes or new features need separate scoping. WS-01 carries written retest criteria: a Northstar member cannot read, modify or delete Cedar records by ID, same-tenant operations continue to work, administrator rules behave as intended, and equivalent checks hold for notes, exports and other contact references.

WS-03 says the retest should confirm execution in a supported browser and assess the deployed Content Security Policy. The remaining findings have no separate retest criteria; the checks sit inside their remediation steps instead.

What to look for in retest reporting

WSTG v4.2 suggests that a re-test report include a subsection summarising the findings of the previous test, the updated status of previously identified vulnerabilities, and any cross-references with the current test. Ask a provider how an updated report shows a finding that changed status, and how it shows one nobody looked at again.

The lab example shows the status vocabulary and the open state. It does not show a finding moving from open to fixed, and it should not be read as if it did.

Negative results, clean-up and the report's own limits

Checks that found nothing are part of the result, and the lab report lists eight of them with what was observed:

  • Tenant override on POST /api/contacts: the session tenant was enforced and the override was not honoured.
  • Tenant override on GET /api/invoices: the override was ignored, and results stayed scoped to the caller's tenant.
  • GET /api/tenants and GET /api/settings: both returned 404 in the tested context.
  • GET /api/admin/settings as a member: 404, with no settings disclosure demonstrated.
  • Note-title rendering: rendered as React text, with no raw HTML sink identified.
  • Note contact filter: the filter was ignored, but the returned list stayed in the caller's tenant.
  • Cross-tenant contact deletion: not tested, to avoid destroying seeded tenant data.
  • Stored XSS browser execution: not dynamically confirmed during the assessment.

Why the caveat matters more than the list

The report closes the list with a caveat: a negative result applies to the tested request, account and application build, and is not a statement about every variation or route. A provider who reports clean results without that line invites you to repeat them more widely than they were tested, for example in a customer questionnaire.

Note too that two of the eight lines are things not done. Listing them beside the clean results keeps an untested area from passing as a tested one.

The tenant and role test matrix

An appendix pulls the results into one tenant and role test matrix of eight rows. Each row gives the test context, the resource or function targeted, the observed result, and the finding, baseline or negative check it supports. Seven rows start from a Northstar member and one from an unauthenticated request carrying the exposed credential.

A row such as "Northstar member, Cedar contact identifier, read and modification succeeded, WS-01" lets a reader see at a glance which boundaries were tried and from which side, which is also where the matrix shows its limit: no row starts from a Cedar account.

Test data left behind

The report records its clean-up. The modified Cedar contact values and the temporarily elevated Northstar role were restored after testing, and the CSV probe contact was deleted. Two XSS test notes remained to be removed, because the application did not expose a note-deletion route. A report that lists leftovers this way tells the owner what is still in the system and why, which matters most when the target is production.

Where the report draws its own lines

The final section restates what the report is not. It does not provide a complete route inventory, a source-code audit, a compliance opinion or assurance about issues outside the recorded test coverage. Its references to the OWASP WSTG, ASVS 5.0, the Top 10:2025 and CVSS v3.1 describe the approach and classification used, and the report says they do not constitute an OWASP certification or a compliance assessment.

A report that draws these lines makes it harder for anyone, the provider included, to stretch it into a claim it cannot support.

Judging the sample report a provider sends you

Ask any provider for a sample before you buy, and before you compare findings, check what kind of sample it is.

These questions work on any pentest report, including this one.

  • Is it labelled? A sample built on a lab or a deliberately vulnerable application should say so, and one drawn from client work should say it was redacted and approved for sharing.
  • Does it keep at least one finding the tester could not fully prove, with the limit stated? A sample in which every finding is fully proven tells you less about how the provider reports uncertainty.
  • Does every score come with a vector, and where a business rating differs from its CVSS band, is the reason written down?
  • Does the scope list exclusions, the roles and accounts used, and the build that was tested?
  • Are negative results recorded, with the caveat that they cover only what was tried?
  • Is the retest vocabulary defined in advance, including a status for findings that were not retested?
  • Does every finding carry a weakness class and a category, with the basis given where they depart from the publisher's list?

Comparing two providers side by side

Put two samples next to each other and run the yardstick table on each, row by row. Then read one finding from each end to end: description, evidence, rating, fix and retest criteria. The differences that matter tend to show up in evidence, limitations and retest status rather than in the number of findings, because your developers can act on a reproducible finding and have to investigate an unproven one before they can.

Frequently asked questions

An executive summary, the scope and its limitations, a findings summary, and for each finding the evidence, impact, rating and fix, followed by remediation priorities and retest status. The reporting chapter of the OWASP Web Security Testing Guide v4.2 also suggests appendices such as the test methodology, severity and risk rating explanations, and a checklist of the tests conducted. The yardstick table on this page breaks each part into what to look for and what to ask.

No. It is a report on crm.lab.test, a lab application Clever Ops controls, with two seeded test tenants and test data. It is not client work, it is not independent assurance, and it is not evidence about any real business. It is published to show the format, with annotations on what each part does and does not demonstrate.

CVSS measures technical severity, and the business rating weighs the data held, the exposure of the affected function, existing controls and the cost of interruption. WS-01 let a low-privilege user read and modify another tenant's confidential customer records, so the report rates it Critical for the business. Its CVSS 3.1 score stays 8.1 High because the affected resources remain within the CRM application's security authority.

Because the evidence was strong and the limit is stated. WS-03 shows HTML stored verbatim through the notes API and a raw HTML rendering path in the public client bundle, but a browser was not available, so script execution was not visually confirmed. The finding is rated High pending browser confirmation, and the retest is to confirm execution in a supported browser. Leaving it out would hide a likely problem, and calling it confirmed would claim more than was shown.

It should list exclusions and limits in so many words. The lab report excludes denial-of-service, load and volumetric testing; destructive deletion of tenant records; third-party services, network infrastructure and host operating systems; social engineering, physical testing and incident response; source-code review beyond public browser assets; compliance certification; and a complete inventory and test of every route. It also states that each negative result covers only the tested request, account and build.

No. A report records what was tested, what was found and what changed after fixes, as evidence for a decision someone else makes, such as a customer or an auditor. The lab report lists compliance certification, and any statement that the application meets a regulatory standard, as out of scope, and says its OWASP and CVSS references are not a certification or a compliance assessment.

Scope the right test for your application

Tell us what you have built, who uses it and when it needs to launch. You will get a plain-English recommendation: a website audit, a web application test or a referral to a specialist.