roadmap
- 01
- 02
- 03
- 04
- 05
- 06
- 07
- 08
Run ZAP against the applications your team owns and turn what it finds into a pipeline gate you can trust. Covers the proxy and scope, logged-in scans, crawling, scan rules, triage and automation.
232 of 232 questions in your plan: 0 known, 0 in progress, 0 due, 0 skipped, 232 new. 0 cut, outside your plan.
ZAP is rarely a round of its own. It surfaces in the security-testing part of QA and SDET loops, and in DevSecOps and application-security interviews about a pipeline that scans the team's own services, often as a scan report pasted into the call with one question: what does this actually prove? Most questions sit in the gap between a run that finished and a run that tested something. A baseline scan exits clean, yet every page the crawler reached was the login form: how would you have noticed? An active scan against a live site ends in seconds. One rule floods the pipeline with alerts, and lowering its attack strength changes nothing, because strength and alert threshold are separate dials. And since ZAP does not check that you are allowed to scan a host, what should a job require before it starts? The same report reads differently by level. Early on, interviewers want to hear which packaged script only crawls and watches traffic and which one also attacks. From middle level up they ask why a run behaved as it did: which context decided scope, which session method carried the login forward, how the per-rule verdict file became an exit code. Senior candidates are asked to run scanning for other teams: which services get an active scan and where, how a false positive stays closed, and when a finding needs a person rather than another rule.
Two prerequisites open the plan, because a scanner reports in their vocabulary. The [OWASP Top 10](/topics/found-appsec-owasp-top-10) names the classes ZAP's rules look for, and HTTP [header fields](/topics/proto-http-headers), [cookies](/topics/proto-http-cookies) and [authentication schemes](/topics/proto-http-auth-framework) are what ZAP alters when it injects and must carry forward when it scans as a logged-in user. ZAP itself starts with [how the program runs](/topics/sec-owasp-zap-runtime), headless or in a container, then puts its proxy in the request path and defines [contexts](/topics/sec-owasp-zap-traffic-context), which later steps consult to decide what is in scope. Logging in comes before crawling: a session that [expires or is logged out](/topics/be-authz-sessions-lifecycle) mid-run is a common reason an authenticated scan quietly turns anonymous, and [ZAP's login and session settings](/topics/sec-owasp-zap-auth) exist to notice it. Only then is the [request list built](/topics/sec-owasp-zap-discovery), by crawling or by importing an [OpenAPI](/topics/proto-openapi) definition that reaches endpoints a crawler may miss. The [scan engines](/topics/sec-owasp-zap-scan) follow because they test only what that stretch collected. Then the plan turns from producing findings to deciding about them: telling a false positive from a [benign true positive](/topics/sec-security-operations-alert-triage-closing-disposition-categories), and keeping known issues in a [baseline](/topics/found-static-analysis-baselines), come before [ZAP's alert filters and reports](/topics/sec-owasp-zap-findings). Automation follows triage on purpose, since the packaged scripts' rule file and exit codes in [machine-driven runs](/topics/sec-owasp-zap-automation) encode decisions triage has already made. The last module is judgement: how a scan differs from [a pentest or a red team](/topics/sec-security-operations-adversary-emulation-designing-engagement-types), what [written authorisation](/topics/sec-security-operations-adversary-emulation-designing-authorisation-deconfliction) a target needs, and what ZAP's [own safeguards](/topics/sec-owasp-zap-limits-permission) do and do not enforce.
Many candidates know ZAP only as a CI step. With a couple of weeks, narrow importance to must know, leave study order on by frequency, and spend the saved hours between logged-in access and automation, where follow-ups about a clean run that tested nothing usually land. QA engineers fluent in HTTP can skim the opening modules; developers should not skip the scan engines, because the passive and active split explains most surprising results. Senior AppSec and DevSecOps candidates should give the final module real time: who signed off a scan, and what protect mode would have stopped, comes up when the role owns scanning for other teams. Practise only against targets you own or are authorised to test, such as a deliberately vulnerable app on your own machine. The plan teaches the scanner, not the whole security role: threat modeling, secrets and supply-chain security are in the [DevSecOps Engineer roadmap](/roadmaps/devsecops), and when ZAP is one round among UI, API and performance testing, the [QA Engineer roadmap](/roadmaps/qa-engineer) carries the rest of that loop.
ZAP is rarely a round of its own. It surfaces in the security-testing part of QA and SDET loops, and in DevSecOps and application-security interviews about a pipeline that scans the team's own services, often as a scan report pasted into the call with one question: what does this actually prove?
Most questions sit in the gap between a run that finished and a run that tested something. A baseline scan exits clean, yet every page the crawler reached was the login form: how would you have noticed? An active scan against a live site ends in seconds. One rule floods the pipeline with alerts, and lowering its attack strength changes nothing, because strength and alert threshold are separate dials. And since ZAP does not check that you are allowed to scan a host, what should a job require before it starts?
The same report reads differently by level. Early on, interviewers want to hear which packaged script only crawls and watches traffic and which one also attacks. From middle level up they ask why a run behaved as it did: which context decided scope, which session method carried the login forward, how the per-rule verdict file became an exit code. Senior candidates are asked to run scanning for other teams: which services get an active scan and where, how a false positive stays closed, and when a finding needs a person rather than another rule.
Two prerequisites open the plan, because a scanner reports in their vocabulary. The OWASP Top 10 names the classes ZAP's rules look for, and HTTP header fields, cookies and authentication schemes are what ZAP alters when it injects and must carry forward when it scans as a logged-in user.
ZAP itself starts with how the program runs, headless or in a container, then puts its proxy in the request path and defines contexts, which later steps consult to decide what is in scope. Logging in comes before crawling: a session that expires or is logged out mid-run is a common reason an authenticated scan quietly turns anonymous, and ZAP's login and session settings exist to notice it. Only then is the request list built, by crawling or by importing an OpenAPI definition that reaches endpoints a crawler may miss.
The scan engines follow because they test only what that stretch collected. Then the plan turns from producing findings to deciding about them: telling a false positive from a benign true positive, and keeping known issues in a baseline, come before ZAP's alert filters and reports. Automation follows triage on purpose, since the packaged scripts' rule file and exit codes in machine-driven runs encode decisions triage has already made.
The last module is judgement: how a scan differs from a pentest or a red team, what written authorisation a target needs, and what ZAP's own safeguards do and do not enforce.
Many candidates know ZAP only as a CI step. With a couple of weeks, narrow importance to must know, leave study order on by frequency, and spend the saved hours between logged-in access and automation, where follow-ups about a clean run that tested nothing usually land. QA engineers fluent in HTTP can skim the opening modules; developers should not skip the scan engines, because the passive and active split explains most surprising results. Senior AppSec and DevSecOps candidates should give the final module real time: who signed off a scan, and what protect mode would have stopped, comes up when the role owns scanning for other teams.
Practise only against targets you own or are authorised to test, such as a deliberately vulnerable app on your own machine. The plan teaches the scanner, not the whole security role: threat modeling, secrets and supply-chain security are in the DevSecOps Engineer roadmap, and when ZAP is one round among UI, API and performance testing, the QA Engineer roadmap carries the rest of that loop.