Skip to main content
Security templateReviewed

Penetration test remediation plan: a tracker from report to retest

A free remediation tracker for the work between a penetration test report and its retest.

Each row ties a finding to its route and role, root cause, owner, fix branch, deployment and evidence. A decision not to fix stays on the record as an open risk with a named owner, a reason and a review date, rather than as a resolved finding.

Remediation tracker with a worked access-control example

Copy the table into a spreadsheet or issue tracker, keep the column headings, and fill in the first row's prompts in round brackets once for each finding. Keep the report's finding IDs.

How to use it

Give a root cause behind several findings its own RC row, and give anything found during remediation its own S ID. The Status column takes one state your team sets (open, in progress, fix deployed, ready for retest, deferred or risk accepted) or, after a retest, the tester's outcome (fixed, partially fixed, still reproducible or not retested). The last column is the management view: count the rows by it and report each count on its own line, never as one total. Every row below the first is an illustration written for this page, not a client's result. portal.training-demo.test is an invented course portal for a training business, with learner, trainer and office roles, and its people, accounts, routes, releases and decisions are invented too. Findings F-01 and F-03 in the invented report have other causes and are left out. Target and review dates are left as prompts because the owner sets them.

8 rows

Remediation tracker with a worked access-control example

Finding or itemAffected route and roleRoot causeOwner, order and target dateFix branch and deploymentEvidence before retestStatusDecision recordManagement view
Blank row to copy: (F-nn, as the report numbers it) (the report's title, unchanged)(method and path of each affected route) (role and test account that showed it) (roles and routes the report says held, or were not checked)(RC-nn) (the missing or faulty control, in one sentence) (confirmed in code, or inferred from behaviour)(one named person) (the report's business rating and CVSS score with its vector, plus your team's rating and reason if it differs) (position in the order of work) (target date the owner commits to)(branch or pull request) (build or release id) (environment) (deployment date)(regression test name and its result on the build) (the report's steps replayed by your team, and the result) (related routes checked) (production review, if the finding called for one) (test data removed)(open, in progress, fix deployed, ready for retest, deferred or risk accepted; after a retest, the tester's outcome and the build)(fix, defer or accept) (who decided, and who approved the change) (reason) (interim or compensating measure) (review date for a deferral or acceptance) (the latest retest outcome and build, if there has been one)(confirmed fixed, fix in hand, deferred, accepted risk, or found during remediation)
RC-1, the root cause behind F-02, F-04, F-05 and S-01. One piece of work, not a finding, so it has no rating and no retest outcome of its own.Every API route outside /api/admin/. Routes named in the report: GET /api/reports/enrolments.csv, POST /api/certificates/issue and POST /api/courses/{courseId}/announcements.The middleware checks for the office role only on routes under /api/admin/, and admits any signed-in user to every other API route. Staff features added later were built outside that prefix and inherited the weaker check. Inferred by the tester from behaviour; confirmed in code by the owner before the fix was designed.T. Nguyen, lead developer (invented). First in the order: one change closes three report findings, all rated business High or Critical. Target date: (set by T. Nguyen).Branch fix/rc-1-deny-by-default, pull request 311. Every API route now declares the permission it needs, the middleware refuses any route that declares none, and refused requests are logged. Deployed to staging as release 2026.12.2 on 9 December 2026.A route inventory test lists all 47 API routes and fails the build if any route declares no permission; passing on release 2026.12.2. Building the inventory turned up one route the report did not name, logged as S-01.Fix deployed. Each linked finding carries its own status.Fix. Approved by R. Haddad, operations manager (invented), before release, as a change to every API route.Not counted. Its findings are counted one by one.
F-02, Learner role can download every learner's enrolment record, with contact details, as a CSV fileGET /api/reports/enrolments.csv. Shown with lrn-test-1 (learner). The report adds that trn-test-1 (trainer) also received the file, and that signed-out requests got 401.RC-1.T. Nguyen, through RC-1. Report rating: business High. Order: first, with RC-1. Target date: as for RC-1.Through RC-1: release 2026.12.2 on staging.Replayed on release 2026.12.2: lrn-test-1 and trn-test-1 now receive 403, and office-test-1 still downloads the file. Regression test "enrolment export by role" passing. No test data to remove.Ready for retest on release 2026.12.2.Fix. Interim measure from the day the report arrived: the route was switched off, and the office used the export under /api/admin/, which already checked the office role.Fix in hand. Not counted as fixed until the tester records it.
F-04, Learner role can issue a completion certificate, including for their own enrolmentPOST /api/certificates/issue. Shown with lrn-test-1 issuing a certificate for its own test enrolment. Not checked, per the report: the trainer role on this route.RC-1.T. Nguyen, through RC-1; production review by R. Haddad. Report rating: business Critical, because a certificate for a course never completed could be shown to an employer. Order: first, with RC-1.Through RC-1: release 2026.12.2 on staging.Replayed on release 2026.12.2: lrn-test-1 receives 403, trn-test-1 receives 403 (the business confirmed trainers should not issue certificates), and office-test-1 can still issue one for a completed test enrolment. Production review: certificates issued by accounts without the office role listed by query; none found, with the query and result attached. The test certificate from the original test has been revoked.Ready for retest on release 2026.12.2.Fix. Interim measure: the route was switched off, and the office issued certificates from the admin screen until the release.Fix in hand.
F-05, Learner role can email an announcement to every learner on a coursePOST /api/courses/{courseId}/announcements. Shown with lrn-test-1 on a test course whose only learners were test accounts, so no real person received a message.RC-1.T. Nguyen, through RC-1. Report rating: business High, because the message comes from the business's own address and could carry a link to a fake sign-in page. Order: first, with RC-1.Through RC-1: release 2026.12.2 on staging.Replayed on release 2026.12.2: lrn-test-1 receives 403, and trn-test-1, who teaches the test course, can still send. Outstanding: the mail queue record showing nothing was queued for the refused request.Fix deployed. Not yet ready for retest, because one evidence item is outstanding.Fix.Fix in hand.
S-01, found during remediation and not a report finding: the attendance export was open to any signed-in userGET /api/reports/attendance.csv. Found by the route inventory test. Before the fix, lrn-test-1 received the file on staging.RC-1.T. Nguyen, through RC-1. Rated by the business, not the tester: High, on the same reasoning as F-02. Order: with RC-1.Through RC-1: release 2026.12.2 on staging.Replayed on release 2026.12.2: lrn-test-1 receives 403, and office-test-1 still downloads the file. Covered by the route inventory test.Fix deployed. Raised with the tester, and checked in the retest round only if both sides agree to add it.Fix, through RC-1.Found during remediation, counted on its own line.
F-06, Trainer role can change assessment marks after a course has closedPATCH /api/assessments/{assessmentId}/marks. Shown with trn-test-1 on a closed test course. The report notes that office-test-1 can do the same, which the business intends.RC-2: the server checks that the trainer teaches the course but never reads the course's closing date, so the rule that marks freeze at closing has no server check. Confirmed in code. Separate from RC-1.T. Nguyen. Report rating: business Medium, because a change needs a trainer account on that course. Order: after RC-1. Target release: (set by T. Nguyen).Branch fix/rc-2-course-closure opened, not deployed. The fix adds a closed state to the course record and needs a data migration.None yet.Deferred. Open.Defer. Decided by R. Haddad with T. Nguyen: the data migration belongs in the next planned release, not the RC-1 release. Interim measure: before certificates are issued for a course, the office checks its mark history for changes made after the closing date. Review: whenever the target release moves.Deferred, open. Shown apart from fixes, with its interim measure.
F-07, Trainer role can read the contact details of learners in courses the trainer does not teachGET /api/learners/{learnerId}/contact. Shown with trn-test-1 reading lrn-test-2, who is enrolled only in another trainer's course.RC-3: contact details are scoped to the business, not to the course. This matches how the business runs today, so the decision is about the rule rather than the code. Confirmed in code.Decision owner: R. Haddad, who owns learner records for the business. Report rating: business Low.No fix branch. Related change in RC-1: trainers can no longer export contact details in bulk, so lookups happen one learner at a time, and each trainer lookup is now logged (release 2026.12.2).Decision record DR-07, signed by R. Haddad and attached. Log entries for trainer lookups seen on staging after release 2026.12.2.Risk accepted. Open. Listed in the retest handover, so the tester marks it not retested with this decision as the reason.Accept. Reason: trainers cover each other's classes at short notice and need to reach those learners. Residual risk: a trainer could look up learners they never teach. Compensating measures: no bulk export for trainers, and every lookup logged with the trainer and the learner. Review date: (set by R. Haddad), and sooner after a complaint, a lookup pattern in the logs with no class cover behind it, or a change to how cover is arranged.Accepted risk, open. Not counted as fixed, closed or resolved.

Finding or item

Affected route and role

  1. Blank row to copy: (F-nn, as the report numbers it) (the report's title, unchanged)
    Affected route and role
    (method and path of each affected route) (role and test account that showed it) (roles and routes the report says held, or were not checked)
    Root cause
    (RC-nn) (the missing or faulty control, in one sentence) (confirmed in code, or inferred from behaviour)
    Owner, order and target date
    (one named person) (the report's business rating and CVSS score with its vector, plus your team's rating and reason if it differs) (position in the order of work) (target date the owner commits to)
    Fix branch and deployment
    (branch or pull request) (build or release id) (environment) (deployment date)
    Evidence before retest
    (regression test name and its result on the build) (the report's steps replayed by your team, and the result) (related routes checked) (production review, if the finding called for one) (test data removed)
    Status
    (open, in progress, fix deployed, ready for retest, deferred or risk accepted; after a retest, the tester's outcome and the build)
    Decision record
    (fix, defer or accept) (who decided, and who approved the change) (reason) (interim or compensating measure) (review date for a deferral or acceptance) (the latest retest outcome and build, if there has been one)
    Management view
    (confirmed fixed, fix in hand, deferred, accepted risk, or found during remediation)
  2. RC-1, the root cause behind F-02, F-04, F-05 and S-01. One piece of work, not a finding, so it has no rating and no retest outcome of its own.
    Affected route and role
    Every API route outside /api/admin/. Routes named in the report: GET /api/reports/enrolments.csv, POST /api/certificates/issue and POST /api/courses/{courseId}/announcements.
    Root cause
    The middleware checks for the office role only on routes under /api/admin/, and admits any signed-in user to every other API route. Staff features added later were built outside that prefix and inherited the weaker check. Inferred by the tester from behaviour; confirmed in code by the owner before the fix was designed.
    Owner, order and target date
    T. Nguyen, lead developer (invented). First in the order: one change closes three report findings, all rated business High or Critical. Target date: (set by T. Nguyen).
    Fix branch and deployment
    Branch fix/rc-1-deny-by-default, pull request 311. Every API route now declares the permission it needs, the middleware refuses any route that declares none, and refused requests are logged. Deployed to staging as release 2026.12.2 on 9 December 2026.
    Evidence before retest
    A route inventory test lists all 47 API routes and fails the build if any route declares no permission; passing on release 2026.12.2. Building the inventory turned up one route the report did not name, logged as S-01.
    Status
    Fix deployed. Each linked finding carries its own status.
    Decision record
    Fix. Approved by R. Haddad, operations manager (invented), before release, as a change to every API route.
    Management view
    Not counted. Its findings are counted one by one.
  3. F-02, Learner role can download every learner's enrolment record, with contact details, as a CSV file
    Affected route and role
    GET /api/reports/enrolments.csv. Shown with lrn-test-1 (learner). The report adds that trn-test-1 (trainer) also received the file, and that signed-out requests got 401.
    Root cause
    RC-1.
    Owner, order and target date
    T. Nguyen, through RC-1. Report rating: business High. Order: first, with RC-1. Target date: as for RC-1.
    Fix branch and deployment
    Through RC-1: release 2026.12.2 on staging.
    Evidence before retest
    Replayed on release 2026.12.2: lrn-test-1 and trn-test-1 now receive 403, and office-test-1 still downloads the file. Regression test "enrolment export by role" passing. No test data to remove.
    Status
    Ready for retest on release 2026.12.2.
    Decision record
    Fix. Interim measure from the day the report arrived: the route was switched off, and the office used the export under /api/admin/, which already checked the office role.
    Management view
    Fix in hand. Not counted as fixed until the tester records it.
  4. F-04, Learner role can issue a completion certificate, including for their own enrolment
    Affected route and role
    POST /api/certificates/issue. Shown with lrn-test-1 issuing a certificate for its own test enrolment. Not checked, per the report: the trainer role on this route.
    Root cause
    RC-1.
    Owner, order and target date
    T. Nguyen, through RC-1; production review by R. Haddad. Report rating: business Critical, because a certificate for a course never completed could be shown to an employer. Order: first, with RC-1.
    Fix branch and deployment
    Through RC-1: release 2026.12.2 on staging.
    Evidence before retest
    Replayed on release 2026.12.2: lrn-test-1 receives 403, trn-test-1 receives 403 (the business confirmed trainers should not issue certificates), and office-test-1 can still issue one for a completed test enrolment. Production review: certificates issued by accounts without the office role listed by query; none found, with the query and result attached. The test certificate from the original test has been revoked.
    Status
    Ready for retest on release 2026.12.2.
    Decision record
    Fix. Interim measure: the route was switched off, and the office issued certificates from the admin screen until the release.
    Management view
    Fix in hand.
  5. F-05, Learner role can email an announcement to every learner on a course
    Affected route and role
    POST /api/courses/{courseId}/announcements. Shown with lrn-test-1 on a test course whose only learners were test accounts, so no real person received a message.
    Root cause
    RC-1.
    Owner, order and target date
    T. Nguyen, through RC-1. Report rating: business High, because the message comes from the business's own address and could carry a link to a fake sign-in page. Order: first, with RC-1.
    Fix branch and deployment
    Through RC-1: release 2026.12.2 on staging.
    Evidence before retest
    Replayed on release 2026.12.2: lrn-test-1 receives 403, and trn-test-1, who teaches the test course, can still send. Outstanding: the mail queue record showing nothing was queued for the refused request.
    Status
    Fix deployed. Not yet ready for retest, because one evidence item is outstanding.
    Decision record
    Fix.
    Management view
    Fix in hand.
  6. S-01, found during remediation and not a report finding: the attendance export was open to any signed-in user
    Affected route and role
    GET /api/reports/attendance.csv. Found by the route inventory test. Before the fix, lrn-test-1 received the file on staging.
    Root cause
    RC-1.
    Owner, order and target date
    T. Nguyen, through RC-1. Rated by the business, not the tester: High, on the same reasoning as F-02. Order: with RC-1.
    Fix branch and deployment
    Through RC-1: release 2026.12.2 on staging.
    Evidence before retest
    Replayed on release 2026.12.2: lrn-test-1 receives 403, and office-test-1 still downloads the file. Covered by the route inventory test.
    Status
    Fix deployed. Raised with the tester, and checked in the retest round only if both sides agree to add it.
    Decision record
    Fix, through RC-1.
    Management view
    Found during remediation, counted on its own line.
  7. F-06, Trainer role can change assessment marks after a course has closed
    Affected route and role
    PATCH /api/assessments/{assessmentId}/marks. Shown with trn-test-1 on a closed test course. The report notes that office-test-1 can do the same, which the business intends.
    Root cause
    RC-2: the server checks that the trainer teaches the course but never reads the course's closing date, so the rule that marks freeze at closing has no server check. Confirmed in code. Separate from RC-1.
    Owner, order and target date
    T. Nguyen. Report rating: business Medium, because a change needs a trainer account on that course. Order: after RC-1. Target release: (set by T. Nguyen).
    Fix branch and deployment
    Branch fix/rc-2-course-closure opened, not deployed. The fix adds a closed state to the course record and needs a data migration.
    Evidence before retest
    None yet.
    Status
    Deferred. Open.
    Decision record
    Defer. Decided by R. Haddad with T. Nguyen: the data migration belongs in the next planned release, not the RC-1 release. Interim measure: before certificates are issued for a course, the office checks its mark history for changes made after the closing date. Review: whenever the target release moves.
    Management view
    Deferred, open. Shown apart from fixes, with its interim measure.
  8. F-07, Trainer role can read the contact details of learners in courses the trainer does not teach
    Affected route and role
    GET /api/learners/{learnerId}/contact. Shown with trn-test-1 reading lrn-test-2, who is enrolled only in another trainer's course.
    Root cause
    RC-3: contact details are scoped to the business, not to the course. This matches how the business runs today, so the decision is about the rule rather than the code. Confirmed in code.
    Owner, order and target date
    Decision owner: R. Haddad, who owns learner records for the business. Report rating: business Low.
    Fix branch and deployment
    No fix branch. Related change in RC-1: trainers can no longer export contact details in bulk, so lookups happen one learner at a time, and each trainer lookup is now logged (release 2026.12.2).
    Evidence before retest
    Decision record DR-07, signed by R. Haddad and attached. Log entries for trainer lookups seen on staging after release 2026.12.2.
    Status
    Risk accepted. Open. Listed in the retest handover, so the tester marks it not retested with this decision as the reason.
    Decision record
    Accept. Reason: trainers cover each other's classes at short notice and need to reach those learners. Residual risk: a trainer could look up learners they never teach. Compensating measures: no bulk export for trainers, and every lookup logged with the trainer and the learner. Review date: (set by R. Haddad), and sooner after a complaint, a lookup pattern in the logs with no class cover behind it, or a change to how cover is arranged.
    Management view
    Accepted risk, open. Not counted as fixed, closed or resolved.

What a penetration test remediation plan covers

A penetration test report says what is wrong.

A penetration test remediation plan says what happens next: who changes what, in what order, how each change will be shown to work, and which findings the business has decided not to fix for now. It covers the time between receiving the report and asking for the retest, and it is the record your developers, your manager and anyone who later asks about the test will read from.

The UK National Cyber Security Centre (NCSC) is plain about who owns that work. Its penetration testing guidance says the test team will have rated each issue and given a potential solution, but that risk assessment and decisions on the application of fixes are your responsibility.

It adds that vulnerability risk assessment and mitigation is a business process and should not be wholly outsourced to the test team. A remediation plan is where those decisions get written down.

NIST's Technical Guide to Information Security Testing and Assessment, SP 800-115, calls the equivalent document a plan of action and milestones (POA&M). Its glossary describes one as a document that identifies tasks needing to be accomplished, with the resources required, any milestones for meeting the tasks and the scheduled milestone completion dates, and the guide says a POA&M will ensure individual vulnerabilities are addressed with specific, measurable, attainable, realistic and tangible actions.

The guide dates from 2008, but the shape carries over to a web application maintained by one developer: one row per finding, with a task, a person and a date on each.

A Clever Ops report already includes a remediation plan in priority order, and each finding carries a fix written for your stack and the criteria the retest will use. The tracker on this page is where that plan becomes assigned work.

It keeps the report's finding IDs, and it stops when every fix in the round sits on one identified build with its evidence attached. From that point the retest belongs to the tester, and the retest checklist covers it.

Turning the report into tracker rows

Start the tracker the day the report arrives, before anyone opens a code editor.

Give each finding one row, and copy across the report's finding ID, title, affected routes and roles, and ratings without rewording them. The ID is what connects your fix to the tester's retest entry later, and a reworded title makes that connection harder to follow.

Where the report lists roles and routes that were checked and held, copy those too, because they show the developer where a working control already exists.

Then read each rating against what you know about the business. NCSC notes that the test team may not have had access to all the details of a system, or to the potential business impact of a vulnerability being exploited, so it may rate issues lower or higher than you would.

It also gives the caution worth writing into the tracker's rules: assessing vulnerability levels should not be used to downplay issues, but to identify the risk to your organisation. If your team changes a rating, the row keeps the report's rating beside yours, with the reason.

Before assigning work, settle anything unclear in the evidence. NCSC suggests a debriefing at which the test team runs through its findings and you can ask for further information or clarification. Bring questions about root causes, because they decide how wide each fix has to be.

One named owner for each row

Each row gets one named person as its owner. A team name gives nobody in particular a reason to pick the row up, and when the person who wrote the code has moved on, the row shows who holds it now.

The owner can be your own developer, your agency's developer, or someone working in a Clever Ops security fix sprint, which covers remediation of known findings, from our report or from another tester's, under its own statement of work.

Keeping the fix work separate from the test means the test result does not depend on who you hire to act on it.

Fix the cause, not the request in the evidence

The evidence in a finding is one request that worked when it should not have.

The root cause is the missing or faulty control that let it work, and one cause can sit behind several findings. SP 800-115 makes the general point in its section on analysis: individual vulnerabilities need to be identified and resolved, but identifying their root cause is key to improving an organisation's overall security posture, because a root cause can often be traced to program-level weaknesses.

Its list of common root causes includes poor integration of security into the system development life cycle, such as missing or unsatisfied security requirements and vulnerabilities in application code the organisation developed.

The worked example in the tracker shows the pattern. Three findings in one invented report, a learner downloading the enrolment export, a learner issuing a course certificate and a learner emailing everyone on a course, look like three bugs on three routes.

They share one cause, logged as RC-1: the application checks for the office role only on routes under /api/admin/, and admits any signed-in user to every other API route.

Staff features added later were built outside that prefix and inherited the weaker check. Patching the three routes would close three findings and leave in place the rule that produced them.

The OWASP Top 10:2025 entry for broken access control names this failure. Among common access control vulnerabilities it lists a violation of the principle of least privilege, commonly known as deny by default, where access should only be granted for particular capabilities, roles or users but is available to anyone.

Its prevention advice includes denying by default, except for public resources, and implementing access control mechanisms once and reusing them throughout the application. The fix for RC-1 follows both: every route declares the permission it needs, and the middleware refuses any route that declares none.

ASVS 5.0 points the same way. Requirement 8.2.1 asks that function-level access is restricted to consumers with explicit permissions, and requirement 8.1.1 asks for authorisation documentation that defines the rules for restricting function-level and data-specific access based on consumer permissions and resource attributes. A route inventory that names each route's permission is one way to write those rules down.

How the tracker keeps a shared cause visible

A shared cause disappears when it lives only in one developer's head. These rules keep it on the page.

  • The root cause gets its own row and ID, such as RC-1, holding the owner, the fix branch and the deployment. It has no rating and no retest outcome, because it is not a finding.
  • Each finding keeps its own row and points to its cause. One change can close several findings, but the tester gives each finding its own outcome, so a route the change missed still shows up.
  • Routes found while fixing the cause, which the report never named, get IDs of their own (S-01 in the example) and their own line in the management view. They are not report findings, so the retest covers them only if you and the tester agree to add them.
  • A finding with two causes points to both, and it is not ready for retest until both are dealt with.
  • The cause row records whether the cause was confirmed in code or inferred from behaviour. A test without code access can only infer it, so the owner confirms it before designing the fix.

A technical change and a process change

SP 800-115 says mitigation recommendations, including the outcome of the root cause analysis, should be developed for each finding, and that there may be both technical recommendations and nontechnical ones that address the organisation's processes.

For RC-1, the technical change is the deny-by-default rule. The process change is a route inventory test that fails the build when a new route declares no permission. The OWASP entry also says developers and QA staff should include functional access control in their unit and integration tests, and that access control failures should be logged, with admins alerted when appropriate, such as after repeated failures.

NCSC asks for the same habit from the other direction. Of vulnerability identification in bespoke or niche software, a type of testing it says is most commonly used in web applications, it says the testing must give feedback to developers on coding practices which avoid introducing the categories of vulnerability identified.

It also says any vulnerability the test identified that you did not previously know about should be given special attention, with the aim of identifying ways you might spot such issues in future.

The matching NIST CSF 2.0 outcome is ID.IM-02: improvements are identified from security tests and exercises, including those done in coordination with suppliers and relevant third parties.

Setting the order of work

The report's priority order is the starting point, not the finished plan.

The plan also has to weigh what a tester cannot see: release dates, who is available, and which fixes depend on others. NIST CSF 2.0 frames this step as two outcomes in its Risk Assessment category. Under ID.RA-05, threats, vulnerabilities, likelihoods and impacts are used to understand inherent risk and inform risk response prioritisation.

Under ID.RA-06, risk responses are chosen, prioritised, planned, tracked and communicated. The CSF describes outcomes and does not prescribe how to achieve them, so the ordering rules below are this template's own.

  • Start with any finding that anyone on the internet, or anyone who can register an account, can reach, where it exposes other people's records, money or control of the application.
  • Put a root cause behind several findings ahead of single findings with the same rating, because one change retires several rows.
  • Move up a finding in a third-party component listed in CISA's Known Exploited Vulnerabilities catalogue. CISA says organisations should use the catalogue as an input to their vulnerability management prioritisation framework.
  • Give each finding that will wait for a later release an interim measure, such as switching a route off, narrowing a role or watching the affected records, and record it on the row.
  • Where a fix needs a data model change or a larger release, record a deferral with its own target instead of letting the date slip without a decision.

Dates belong to the owner

Each owner sets the target date for their own rows after reading the finding, and the tracker records it as their commitment. A tester can recommend an order of work, but only the person doing the work can commit to a date that fits your releases, so the template leaves every target date blank until the owner fills it in.

Approval matters for changes like RC-1, which touches every API route. SP 800-115 asks for the plan to be coordinated through the organisation's configuration control or configuration management board, because it is likely to propose changes to existing systems, networks, policy or processes, and says that at a minimum the program manager or system owner should approve the planned mitigation actions before they are implemented.

In a small business the approver might be the director or the person who owns the product. The tracker's decision column records who approved and when.

How a fix is evidenced before the retest

A developer saying a finding is fixed is a claim about their own work.

The evidence column is what lets someone else check that claim, and it is where the tester starts. For each row, record the following.

  • The branch or pull request, and the commit or release that carries the change.
  • The deployment record: the environment, the build, and the date it went live there.
  • The regression test that encodes the finding, by name, with its result on that build. For an access-control finding it sends the request as each role that should be refused and each role that should still succeed.
  • The report's reproduction steps, replayed by your team on that build with the test accounts the tester used, and the result.
  • Related routes and roles your team checked, starting with any the report listed as not checked.
  • Any production review the finding called for, such as listing the records created through the flawed route, and what it found.
  • Test data from the original test: removed, or still listed for removal.

Try the fix where it cannot hurt anyone

SP 800-115 puts a step before production. Before technical changes are made to a production asset, it says, they should be tested on systems in an environment that replicates the one where the change will be deployed, and it notes that such testing significantly reduces, but does not eliminate, the risk of a system reacting adversely to the change.

For a web application the equivalent is a staging copy, with the regression tests running there before the release goes out.

Why deployed is not the same as fixed

SP 800-115 lists three ways to verify that mitigation actions were implemented: an audit of the system, retesting the system and its components, and holding personnel accountable through documentation. It notes that the test team can verify an implementation only if a mirror copy of the original test is performed, and that while verification by signed documentation is more cost-effective in the short term, there are risks in not technically verifying that changes have been made.

That is why the tracker keeps the developer's states apart from the tester's outcomes. Fix deployed and ready for retest are your team's states. Fixed is the tester's call, recorded after the retest repeats the original steps on the named build.

Status words: six before the retest, four after

Each row carries one status at a time.

Your team sets the first six as the work moves. The last four are the retest outcomes used throughout this library, and only the tester writes them.

  • Open: an owner is named and the root cause is recorded as confirmed or inferred. No change has started.
  • In progress: a branch or pull request exists for the fix.
  • Fix deployed: the change is on a named build, and the evidence is still being gathered.
  • Ready for retest: the change is on the build the retest will use, and every evidence item for the row is attached.
  • Deferred: the fix is planned for a later release, with a target the owner sets and an interim measure. The finding stays open.
  • Risk accepted: a named owner has decided not to fix the finding for now, with a reason and a review date. The finding stays open.
  • Fixed, partially fixed, still reproducible or not retested: the tester's outcome for the round, with the build it was checked on. Fixed means the original reproduction fails and the agreed regression checks pass. Partially fixed means some affected routes, or part of the impact, can still be reproduced. Still reproducible means the original issue still works under the agreed test conditions. Not retested means the finding was not checked in that round, and the reason is recorded, for example that it was left out or that the access or build it needed was unavailable.

What happens to a row after the retest

The tester's outcome stays in the record for that round. A row marked partially fixed or still reproducible then goes back to in progress, unless its owner decides to defer it or accept the risk, in which case it moves to that state, and its decision record notes the outcome and build of that round.

A row marked not retested keeps the state it had before the round. No row moves to fixed on the strength of a developer's note, however confident.

Recording a decision not to fix

Some findings will not be fixed in this round, and that can be a sound decision.

NIST CSF 2.0 says an organisation may choose to handle risk in one or more ways, including mitigating, transferring, avoiding or accepting negative risks, depending on the potential impacts and likelihoods. The trouble starts when acceptance is written down as if it were a fix. The CSF outcome to point to is ID.RA-07: changes and exceptions are managed, assessed for risk impact, recorded and tracked.

Deferred and accepted are different decisions

A deferral says the fix will happen later. It carries a target release and an interim measure until then. An acceptance says the business will live with the risk for now. It carries a reason and a review date instead of a fix date.

Both keep the finding open, and neither is shown as fixed, closed or resolved anywhere: not in the management view, and not in answers to a customer.

What the decision record holds

Each deferred or accepted row links to a short decision record. It holds:

  • The finding ID and the report's rating, unchanged.
  • The decision: defer or accept.
  • Who decided: a named person who owns the risk for the business, not the developer assigned to the fix.
  • The reason, in a sentence a customer's security reviewer could read.
  • The residual risk: what someone could still do while the finding stays open.
  • Interim or compensating measures, each with the release or setting that put it in place.
  • A review date the owner sets, and the events that bring the review forward, such as a complaint, a sign of misuse in the logs or a change to the feature.
  • The latest retest outcome and the build it was checked on, if there has been a retest.

Looking past the suggested fix

NCSC's guidance supports weighing other options before deciding. It says the solutions a tester proposes may not be the only ones possible, and that you should take advice from your own technical staff and suppliers on alternatives. Taking a suggested patch as its example, it says it may be possible to uninstall the software if it is not actually required, or to put other controls in place to limit exposure to the vulnerability, and that additional monitoring of the vulnerable component may even be sufficient to reduce the risk to an acceptable level.

Where monitoring is part of the reason a risk is accepted, the decision record says what is monitored and who reads it.

The management view: accepted risk is not a completed fix

Whoever funds the work wants one line: how much is done.

The honest answer has five parts, and the management view keeps them apart rather than adding them into one "resolved" total.

  • Confirmed fixed: findings the tester marked fixed at a retest, with the build. This is the only count of completed fixes.
  • Fix in hand: open, in progress, fix deployed and ready for retest, plus anything a retest marked partially fixed or still reproducible.
  • Deferred, open: each with its target release and interim measure.
  • Accepted risk, open: each with its decision owner and review date.
  • Found during remediation: items the report never named, counted apart so they neither swell nor hide the report's own numbers.

The worked example at handover

When the invented portal's fixes reach staging, its management view reads: confirmed fixed, none yet; fix in hand, three report findings (F-02 and F-04 ready for retest, F-05 deployed with one evidence item outstanding); deferred, one (F-06); accepted risk, one (F-07); found during remediation, one (S-01).

Four rows are deployed on staging and the confirmed count is still zero. Until the retest, that is the accurate position to give a board, a customer or an insurer.

Questionnaires follow the same rule. A question asking whether all findings were remediated needs each finding's retest outcome, and a deferred or accepted finding is answered as open, with the decision behind it.

Handing the plan over for the retest

The plan ends when the rows you intend to retest are ready.

For a Clever Ops test, you request the one consolidated retest within the window set in your statement of work, once the fixes are deployed to an identified build of the same application. Whoever runs your retest, check these before you ask for it.

  • Every fix in the round is on one build, and each row names that build.
  • The test accounts and test records from the original test still exist, so the tester can replay the original requests.
  • Each row marked ready for retest has its evidence attached.
  • Deferred and accepted findings are listed with their decision records, so the tester can mark each one not retested with your reason instead of leaving a blank.
  • Items found during remediation are flagged, with a question about whether they can be checked in the same round. The retest covers the reported findings within the agreed scope, so adding them is agreed separately.

Where this template stops

From the handover on, the retest is the tester's. Each finding gets one outcome on the named build, and the tracker's status column takes that outcome exactly as the tester wrote it. Rows that come back open return to the top of this page: a new owner decision, a new branch and new evidence before the next round.

Frequently asked questions

One row per finding, keeping the report's ID, with the affected routes and roles, the root cause, a named owner, the order of work, a target date the owner sets, the fix branch and deployment, the evidence your team gathered and a status. Add a row for each root cause behind several findings, a decision record for anything deferred or accepted, and a management view that counts confirmed fixes apart from everything else.

The developer can mark it fix deployed, and ready for retest once the evidence is attached. Fixed is the outcome a retest records after repeating the original steps on a named build. Keeping the two apart stops a board or a customer from being told a finding is closed on the strength of a note.

No. An accepted risk stays open, with a named owner, a reason, any compensating measures and a review date. It is counted under accepted risk in the management view and answered as open in a questionnaire. At the retest it is marked not retested, with the decision as the reason.

The ledger in a report records the tester's outcome for each finding in each retest round. The remediation tracker is your team's working record before that: who owns each finding, the root cause, the branch and build, and the evidence your team gathered. The tester writes the ledger; the business keeps the tracker.

The pattern. The request in the evidence shows the cause, and the fix belongs to the cause. Give the cause its own row, list every route it reaches, and add a regression test that fails if the pattern comes back. Routes your developer finds along the way go into the tracker as items found during remediation.

Your own developers, your agency, or a Clever Ops security fix sprint, which covers remediation of known findings, from our report or another tester's, under its own statement of work. If the report came from a Clever Ops test, one consolidated retest of the reported findings is included, whoever makes the fixes.

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.