You inherit 320 help-center pages with no owner data and have 10 working days before planning; how would you produce a content-audit workbook that supports deletion and rewrite decisions?

I would turn the audit into a decision tool rather than a list of URLs. - I would crawl URL, title, last modified date, traffic, search entrances, inbound links, and owner into one row per page. - I would score accuracy and task value separately, because a low-traffic compliance page may still be mandatory. - I would assign keep, update, merge, archive, or delete with evidence and a named approver for every destructive action. - I would deliver a summary by product area plus a prioritized 30-day backlog, not ask stakeholders to read 320 rows.

Documentation is split across Confluence, GitHub, and a legacy CMS with about 480 pages, and owners discover stale material only through complaints; how would you deliver a lifecycle metadata and ownership schema in two weeks?

I would define lifecycle as an enforceable state machine tied to accountable product domains. - Required metadata would include a stable content ID, lifecycle state, risk tier, product owner, documentation owner, last verified date, review interval, source version, and replacement URL. - Allowed transitions would cover draft, active, review due, stale, deprecated, and archived, with entry criteria, permitted approvers, and required evidence for each change. - Scheduled checks would move active content to review due or stale when verification expires, ownership disappears, a supported version retires, or a source contract changes, while never auto-archiving high-risk pages. - I would pilot the rules across all three systems, report invalid and unowned records, and publish an escalation queue so all 480 pages converge on the schema.

A three-product SaaS suite has 260 pages grouped by internal team names, and product wants a taxonomy proposal in three weeks; how would you deliver a taxonomy decision record and controlled vocabulary?

I would base the taxonomy on user tasks and stable product concepts, not the current org chart. - I would sample support tickets, search queries, interviews, and page analytics to identify the language customers actually use. - I would define a small set of facets such as product, task, persona, and lifecycle, with allowed values and explicit exclusion rules. - I would test candidate labels through tree testing against top tasks before applying them to all 260 pages. - The decision record would include term definitions, governance, migration mapping, and examples of ambiguous pages.

Users need six clicks to reach setup instructions in a 180-page portal, and you have 12 days to deliver a navigation-tree prototype; how would you redesign and validate it?

I would optimize the tree around the highest-value user journeys and preserve escape routes for secondary tasks. - I would identify the top 15 tasks from analytics and support data, then map every current page to one primary journey. - I would keep labels concrete and mutually distinct, limiting each navigation level so scanning does not become another search problem. - I would run tree tests with representative users and track first-click success, path directness, and failures by label. - The prototype would include old-to-new mappings and redirect requirements so approval leads directly to implementation.

The portal logged 1,500 unique queries last month, 22 percent returned no useful result, and you have five days to deliver a search-term and synonym map; what would you include?

I would map user language to canonical concepts and separate vocabulary fixes from missing-content work. - I would cluster queries after normalizing casing, spelling, plurals, product aliases, and common error-code variants. - For each high-volume cluster, I would record the preferred term, safe synonyms, intended landing page, and whether a new page is required. - I would reject broad synonyms that increase recall but mix distinct concepts, then test changes against a fixed relevance set. - The artifact would rank clusters by query volume, zero-result rate, and business-critical tasks for the next release.

Five writers use inconsistent structures across 220 pages, and you have two weeks to deliver page-template specifications for tutorials, how-to guides, concepts, and references; how would you design them?

I would make each template enforce the reader goal while leaving room for domain-specific content. - Every specification would state audience, entry conditions, expected outcome, required sections, optional sections, and prohibited content. - Tutorial templates would preserve a guided learning path, while how-to templates would start from a defined task and avoid teaching every underlying concept. - Reference templates would favor predictable fields and completeness, while concept pages would explain relationships and boundaries. - I would pilot each template on two existing pages and revise it from writer friction before rolling it across 220 pages.

A 400-page knowledge base must feed web, in-product help, and support macros, and engineering wants a content-model specification in three weeks; how would you model it?

I would model reusable meaning at the smallest stable unit without fragmenting prose into unmaintainable fields. - Core entities would include article, procedure, step, prerequisite, warning, code sample, product version, locale, and reusable snippet. - Relationships and validation rules would express where units can appear, which fields are required, and how variants are selected. - I would keep presentation choices out of the schema and test authoring effort as well as channel rendering. - The specification would include JSON examples, cardinality, fallback behavior, migration mappings, and ownership for schema changes.

A 300-page site mixes explanation, procedures, lessons, and parameter tables on the same pages, and you have 15 working days to deliver a Diátaxis separation map; how would you decide what to split?

I would classify sections by the reader's immediate mode, then split only where mixed intent blocks successful use. - Tutorials would provide a controlled learning sequence with a known starting state and a working result. - How-to guides would solve one real task, explanations would build understanding, and references would describe the system consistently and completely. - I would retain short contextual links instead of duplicating conceptual prose inside every procedure. - The map would show current sections, target content type, destination page, cross-links, and migration priority for all mixed pages.

Users complete setup but then lose their path between configuration, verification, and troubleshooting across a 500-page portal; how would you deliver a contextual-linking specification in two weeks?

I would define links as typed transitions in user journeys rather than add generic related-content lists. - I would map the 12 highest-volume journeys and mark where readers need a prerequisite, next step, concept, reference, recovery path, or escalation. - Each link type would have placement, label, destination, and maximum-count rules so a procedure does not end with ten equal-weight choices. - The specification would define reciprocal-link exceptions, version and locale behavior, and checks for dead ends, circular paths, and links into unsupported content. - I would apply it to 30 representative pages and compare next-step clicks and successful journey completion with the current paths.

You must move 350 pages to a new information architecture in six weeks while preserving bookmarked URLs; how would you deliver the migration map and cutover plan?

I would make every source page resolve to an explicit target state before the cutover. - The migration map would assign each URL to move, merge, rewrite, archive, or delete, with a target URL and content owner. - I would generate permanent redirects from the map and test chains, loops, fragments, canonicals, and high-traffic external links. - Content would move in domain batches with acceptance checks for navigation, search indexing, analytics, and access permissions. - The cutover plan would include a freeze window, rollback criteria, monitoring, and a 30-day redirect and 404 review.

A new public API has 120 endpoints in an OpenAPI file and launches in four weeks; how would you deliver a reference section that is accurate but not merely generated output?

I would use OpenAPI as the reference source while adding the human context required to complete real tasks. - I would validate operation IDs, descriptions, schemas, examples, security requirements, and response codes with endpoint owners before generation. - Generated pages would share consistent parameter and response tables, while authored introductions would explain purpose, permissions, side effects, and common sequences. - I would add tested request and response examples for the highest-use workflows rather than fabricate examples for all 120 endpoints. - Release gates would compare the published reference with the OpenAPI artifact and fail on undocumented breaking changes.

Developers must choose between API keys and OAuth 2.0 for a partner API, and you have 10 days to deliver an authentication onboarding quickstart for a 15-minute first call; how would you structure it?

I would provide one shortest secure path per credential type and defer protocol depth until after the first successful call. - The opening decision table would map server-to-server access to scoped API keys and delegated user access to the documented OAuth flow. - Each path would list prerequisites, credential creation, secure storage, required headers or scopes, token expiry, and one copyable request. - I would use non-production credentials, redact secrets in screenshots, and show the expected success and authentication-error responses. - A timed usability test with new developers would verify that the quickstart reaches a valid call within 15 minutes.

The API supports JavaScript, Python, Java, and Go SDKs, and you have three weeks to deliver a runnable examples repository covering eight core workflows; how would you keep examples equivalent and maintainable?

I would define one language-neutral behavior contract and implement it idiomatically in each SDK. - Every workflow would share fixtures, expected output, error cases, and a minimum supported SDK version. - Language examples would follow native conventions for configuration, async calls, resource cleanup, and exception handling instead of line-by-line translation. - CI would install fresh dependencies, inject sandbox credentials, run all 32 examples, and fail when output diverges from the contract. - The repository README would map workflows across languages and name an SDK owner for version updates.

A platform emits 18 webhook event types and partners launch in two weeks; how would you deliver a webhook integration guide that covers signature verification, retries, and testing?

I would organize the guide around receiving one event safely before expanding to the full event catalog. - The first implementation would preserve the raw body, verify the timestamped signature, enforce a replay window, and return a 2xx quickly. - I would document retry timing, duplicate delivery, out-of-order events, and the stable event ID needed for deduplication. - A reference table would define all 18 event types, payload versions, triggering actions, and links to schemas. - The guide would include a sandbox trigger, sample payloads, local tunneling steps, and a failure checklist validated with a partner engineer.

Nine list endpoints use cursor pagination but clients keep treating cursors as page numbers, and you have five days to deliver a pagination pattern page; what would it contain?

I would define the cursor as an opaque continuation token and demonstrate the complete traversal loop. - The page would show the initial request, next-cursor response field, termination condition, maximum page size, and one tested SDK example. - It would state that clients must not parse, modify, cache indefinitely, or assume ordering from the cursor value. - I would explain how inserts, deletes, filters, and sort order affect traversal, including whether duplicates or omissions are possible. - All nine endpoint pages would link to this shared pattern and expose only their endpoint-specific limits.

Six payment endpoints support idempotency keys and go live in one week; how would you deliver a client-facing idempotency guide that prevents duplicate charges?

I would document idempotency as a retry contract with explicit key scope, lifetime, and response behavior. - Clients would generate a unique key per logical operation, reuse it only for retries, and keep the request payload unchanged. - A table would define the 24-hour retention window, concurrent duplicate behavior, status codes, and what happens after key expiry. - Examples would cover a timeout after submission, the safe retry, a mismatched payload, and retrieving the original result. - I would verify wording and examples against integration tests for all six endpoints before publication.

Support has identified 64 public API error codes with inconsistent messages, and you have two weeks to deliver an error catalog plus troubleshooting paths; how would you design it?

I would make every error entry answer what failed, whether retry is safe, and what the caller can do next. - The catalog would record stable code, HTTP status, user-facing meaning, likely causes, retryability, remediation, and a sanitized example response. - I would group by caller action such as fix input, refresh credentials, wait and retry, or contact support rather than by internal service. - Shared guidance would cover request IDs, rate-limit headers, backoff, and the diagnostic details support needs. - Engineering owners would approve machine-readable mappings so prose cannot drift from runtime behavior.

Three products ship weekly and release notes arrive in different formats, and you have five days to deliver a changelog template and intake process; how would you standardize them?

I would standardize entries around user impact while keeping an auditable link to each shipped change. - The template would require date, product, change type, affected audience, availability, action required, and links to updated documentation. - Breaking, deprecated, security-related, fixed, and added changes would use controlled labels with explicit inclusion rules. - Release PRs would supply structured input before the cutoff, with product owners accountable for completeness and writers for clarity. - I would publish one chronological feed with product filters and measure late or missing entries after four release cycles.

API v1 has 220 active customer integrations and will be retired in 90 days, and you have four weeks to deliver a v1-to-v2 migration guide; how would you structure and verify it?

I would make the guide a complete change map with a testable path from one working integration to another. - A compatibility table would list renamed endpoints, schema changes, authentication differences, removed behavior, and exact replacements. - The guide would sequence inventory, sandbox update, dual testing, production cutover, monitoring, and rollback checkpoints. - Before-and-after requests would use realistic payloads and include a script or SDK example for the highest-volume integration pattern. - I would test the guide with two customer engineers and track unresolved migration questions as documentation defects.

A comparison finds 35 differences between the production API and its OpenAPI description, and you have eight days to deliver a reconciliation report before SDK generation; how would you resolve source-of-truth conflicts?

I would classify each difference by production risk and obtain an explicit owner decision instead of silently editing the specification. - I would reproduce mismatches for paths, required fields, enums, defaults, status codes, and nullable behavior against a controlled environment. - The report would record observed runtime behavior, current schema, intended contract, affected clients, owner, and chosen correction. - Safe specification fixes could merge immediately, while runtime changes would require compatibility review and release notes. - SDK generation would remain blocked until critical mismatches close and automated contract checks cover the corrected cases.

Perguntas de entrevista: Technical Writer

100 perguntas reais com respostas modelo e explicações para candidatos Redator Técnico.

Ver um exemplo de currículo: Technical Writer →

Treinar com flashcards

Repetição espaçada · Hunter Pass

Perguntas

content-auditownership

I would turn the audit into a decision tool rather than a list of URLs.

I would crawl URL, title, last modified date, traffic, search entrances, inbound links, and owner into one row per page.
I would score accuracy and task value separately, because a low-traffic compliance page may still be mandatory.
I would assign keep, update, merge, archive, or delete with evidence and a named approver for every destructive action.
I would deliver a summary by product area plus a prioritized 30-day backlog, not ask stakeholders to read 320 rows.

Por que perguntam isso: The interviewer is testing whether the candidate can convert a bounded audit into defensible content decisions and an executable backlog.

documentationcontent-lifecyclecontent-ownership

I would define lifecycle as an enforceable state machine tied to accountable product domains.

Required metadata would include a stable content ID, lifecycle state, risk tier, product owner, documentation owner, last verified date, review interval, source version, and replacement URL.
Allowed transitions would cover draft, active, review due, stale, deprecated, and archived, with entry criteria, permitted approvers, and required evidence for each change.
Scheduled checks would move active content to review due or stale when verification expires, ownership disappears, a supported version retires, or a source contract changes, while never auto-archiving high-risk pages.
I would pilot the rules across all three systems, report invalid and unowned records, and publish an escalation queue so all 480 pages converge on the schema.

Por que perguntam isso: The interviewer is evaluating lifecycle modeling, accountable ownership, automated stale-state transitions, and safe enforcement across heterogeneous systems.

content-taxonomycontrolled-vocabularyforms

I would base the taxonomy on user tasks and stable product concepts, not the current org chart.

I would sample support tickets, search queries, interviews, and page analytics to identify the language customers actually use.
I would define a small set of facets such as product, task, persona, and lifecycle, with allowed values and explicit exclusion rules.
I would test candidate labels through tree testing against top tasks before applying them to all 260 pages.
The decision record would include term definitions, governance, migration mapping, and examples of ambiguous pages.

Por que perguntam isso: The interviewer is checking whether the candidate can derive, validate, and govern a taxonomy instead of merely renaming folders.

validationprototypes

I would optimize the tree around the highest-value user journeys and preserve escape routes for secondary tasks.

I would identify the top 15 tasks from analytics and support data, then map every current page to one primary journey.
I would keep labels concrete and mutually distinct, limiting each navigation level so scanning does not become another search problem.
I would run tree tests with representative users and track first-click success, path directness, and failures by label.
The prototype would include old-to-new mappings and redirect requirements so approval leads directly to implementation.

Por que perguntam isso: The interviewer is testing task-based information architecture, validation methods, and migration awareness.

queries

I would map user language to canonical concepts and separate vocabulary fixes from missing-content work.

I would cluster queries after normalizing casing, spelling, plurals, product aliases, and common error-code variants.
For each high-volume cluster, I would record the preferred term, safe synonyms, intended landing page, and whether a new page is required.
I would reject broad synonyms that increase recall but mix distinct concepts, then test changes against a fixed relevance set.
The artifact would rank clusters by query volume, zero-result rate, and business-critical tasks for the next release.

Por que perguntam isso: The interviewer is evaluating practical search vocabulary design and the ability to balance recall, precision, and content gaps.

tutorialshow-to-guidesreference-docs

I would make each template enforce the reader goal while leaving room for domain-specific content.

Every specification would state audience, entry conditions, expected outcome, required sections, optional sections, and prohibited content.
Tutorial templates would preserve a guided learning path, while how-to templates would start from a defined task and avoid teaching every underlying concept.
Reference templates would favor predictable fields and completeness, while concept pages would explain relationships and boundaries.
I would pilot each template on two existing pages and revise it from writer friction before rolling it across 220 pages.

Por que perguntam isso: The interviewer is checking whether the candidate understands template semantics and can validate templates against real authoring work.

I would model reusable meaning at the smallest stable unit without fragmenting prose into unmaintainable fields.

Core entities would include article, procedure, step, prerequisite, warning, code sample, product version, locale, and reusable snippet.
Relationships and validation rules would express where units can appear, which fields are required, and how variants are selected.
I would keep presentation choices out of the schema and test authoring effort as well as channel rendering.
The specification would include JSON examples, cardinality, fallback behavior, migration mappings, and ownership for schema changes.

Por que perguntam isso: The interviewer is testing structured-content modeling, reuse boundaries, channel independence, and governance.

diataxis

I would classify sections by the reader's immediate mode, then split only where mixed intent blocks successful use.

Tutorials would provide a controlled learning sequence with a known starting state and a working result.
How-to guides would solve one real task, explanations would build understanding, and references would describe the system consistently and completely.
I would retain short contextual links instead of duplicating conceptual prose inside every procedure.
The map would show current sections, target content type, destination page, cross-links, and migration priority for all mixed pages.

Por que perguntam isso: The interviewer is evaluating genuine Diátaxis reasoning rather than superficial labeling of existing pages.

troubleshootingconfig

I would define links as typed transitions in user journeys rather than add generic related-content lists.

I would map the 12 highest-volume journeys and mark where readers need a prerequisite, next step, concept, reference, recovery path, or escalation.
Each link type would have placement, label, destination, and maximum-count rules so a procedure does not end with ten equal-weight choices.
The specification would define reciprocal-link exceptions, version and locale behavior, and checks for dead ends, circular paths, and links into unsupported content.
I would apply it to 30 representative pages and compare next-step clicks and successful journey completion with the current paths.

Por que perguntam isso: The interviewer is checking whether the candidate can model purposeful content relationships and validate a cross-linking architecture against real journeys.

information-architecturemigrationsarchitecture

I would make every source page resolve to an explicit target state before the cutover.

The migration map would assign each URL to move, merge, rewrite, archive, or delete, with a target URL and content owner.
I would generate permanent redirects from the map and test chains, loops, fragments, canonicals, and high-traffic external links.
Content would move in domain batches with acceptance checks for navigation, search indexing, analytics, and access permissions.
The cutover plan would include a freeze window, rollback criteria, monitoring, and a 30-day redirect and 404 review.

Por que perguntam isso: The interviewer is evaluating whether the candidate can connect information architecture decisions to a safe, testable migration.

endpointsopenapireference-docs

I would use OpenAPI as the reference source while adding the human context required to complete real tasks.

I would validate operation IDs, descriptions, schemas, examples, security requirements, and response codes with endpoint owners before generation.
Generated pages would share consistent parameter and response tables, while authored introductions would explain purpose, permissions, side effects, and common sequences.
I would add tested request and response examples for the highest-use workflows rather than fabricate examples for all 120 endpoints.
Release gates would compare the published reference with the OpenAPI artifact and fail on undocumented breaking changes.

Por que perguntam isso: The interviewer is testing source-of-truth discipline and the distinction between generated reference data and authored developer guidance.

authoauthapi

I would provide one shortest secure path per credential type and defer protocol depth until after the first successful call.

The opening decision table would map server-to-server access to scoped API keys and delegated user access to the documented OAuth flow.
Each path would list prerequisites, credential creation, secure storage, required headers or scopes, token expiry, and one copyable request.
I would use non-production credentials, redact secrets in screenshots, and show the expected success and authentication-error responses.
A timed usability test with new developers would verify that the quickstart reaches a valid call within 15 minutes.

Por que perguntam isso: The interviewer is evaluating secure onboarding design, progressive disclosure, and a measurable first-success criterion.

javascriptpythonapi

I would define one language-neutral behavior contract and implement it idiomatically in each SDK.

Every workflow would share fixtures, expected output, error cases, and a minimum supported SDK version.
Language examples would follow native conventions for configuration, async calls, resource cleanup, and exception handling instead of line-by-line translation.
CI would install fresh dependencies, inject sandbox credentials, run all 32 examples, and fail when output diverges from the contract.
The repository README would map workflows across languages and name an SDK owner for version updates.

Por que perguntam isso: The interviewer is checking cross-SDK consistency, idiomatic code, executable validation, and ownership.

webhooks

I would organize the guide around receiving one event safely before expanding to the full event catalog.

The first implementation would preserve the raw body, verify the timestamped signature, enforce a replay window, and return a 2xx quickly.
I would document retry timing, duplicate delivery, out-of-order events, and the stable event ID needed for deduplication.
A reference table would define all 18 event types, payload versions, triggering actions, and links to schemas.
The guide would include a sandbox trigger, sample payloads, local tunneling steps, and a failure checklist validated with a partner engineer.

Por que perguntam isso: The interviewer is evaluating whether webhook documentation teaches operationally safe behavior rather than only listing payloads.

endpointspagination

I would define the cursor as an opaque continuation token and demonstrate the complete traversal loop.

The page would show the initial request, next-cursor response field, termination condition, maximum page size, and one tested SDK example.
It would state that clients must not parse, modify, cache indefinitely, or assume ordering from the cursor value.
I would explain how inserts, deletes, filters, and sort order affect traversal, including whether duplicates or omissions are possible.
All nine endpoint pages would link to this shared pattern and expose only their endpoint-specific limits.

Por que perguntam isso: The interviewer is testing the ability to document pagination semantics, edge cases, and reusable API patterns precisely.

endpointsidempotency

I would document idempotency as a retry contract with explicit key scope, lifetime, and response behavior.

Clients would generate a unique key per logical operation, reuse it only for retries, and keep the request payload unchanged.
A table would define the 24-hour retention window, concurrent duplicate behavior, status codes, and what happens after key expiry.
Examples would cover a timeout after submission, the safe retry, a mismatched payload, and retrieving the original result.
I would verify wording and examples against integration tests for all six endpoints before publication.

Por que perguntam isso: The interviewer is evaluating precise documentation of a high-risk API contract and its retry edge cases.

troubleshootingapidesign

I would make every error entry answer what failed, whether retry is safe, and what the caller can do next.

The catalog would record stable code, HTTP status, user-facing meaning, likely causes, retryability, remediation, and a sanitized example response.
I would group by caller action such as fix input, refresh credentials, wait and retry, or contact support rather than by internal service.
Shared guidance would cover request IDs, rate-limit headers, backoff, and the diagnostic details support needs.
Engineering owners would approve machine-readable mappings so prose cannot drift from runtime behavior.

Por que perguntam isso: The interviewer is checking error-information architecture, actionable troubleshooting, and synchronization with implementation.

release-noteschangelogconcurrency

I would standardize entries around user impact while keeping an auditable link to each shipped change.

The template would require date, product, change type, affected audience, availability, action required, and links to updated documentation.
Breaking, deprecated, security-related, fixed, and added changes would use controlled labels with explicit inclusion rules.
Release PRs would supply structured input before the cutoff, with product owners accountable for completeness and writers for clarity.
I would publish one chronological feed with product filters and measure late or missing entries after four release cycles.

Por que perguntam isso: The interviewer is evaluating whether the candidate can design a sustainable changelog system rather than rewrite notes manually each week.

migration-guidesapimigrations

I would make the guide a complete change map with a testable path from one working integration to another.

A compatibility table would list renamed endpoints, schema changes, authentication differences, removed behavior, and exact replacements.
The guide would sequence inventory, sandbox update, dual testing, production cutover, monitoring, and rollback checkpoints.
Before-and-after requests would use realistic payloads and include a script or SDK example for the highest-volume integration pattern.
I would test the guide with two customer engineers and track unresolved migration questions as documentation defects.

Por que perguntam isso: The interviewer is testing migration-document design, completeness, practical validation, and customer risk reduction.

openapireact

I would classify each difference by production risk and obtain an explicit owner decision instead of silently editing the specification.

I would reproduce mismatches for paths, required fields, enums, defaults, status codes, and nullable behavior against a controlled environment.
The report would record observed runtime behavior, current schema, intended contract, affected clients, owner, and chosen correction.
Safe specification fixes could merge immediately, while runtime changes would require compatibility review and release notes.
SDK generation would remain blocked until critical mismatches close and automated contract checks cover the corrected cases.

Por que perguntam isso: The interviewer is evaluating disciplined contract reconciliation, ownership, and protection of generated consumers.

Perguntas bloqueadas

21
You must move 450 Markdown pages into GitHub and launch a Docusaurus site in six weeks for 25 contributors; how would you deliver the docs-as-code architecture decision record?
docs-as-codemarkdowndocusaurus
22
A team of six is choosing between MkDocs Material and Docusaurus for a 180-page Python-heavy portal, and you have 10 days to deliver a recommendation report; how would you run the evaluation?
docusauruspython
23
Thirty contributors produce inconsistent terminology, and you have two weeks to deliver a Vale configuration mapped to the existing style guide; how would you introduce it without blocking useful contributions?
configvalestyle-guides
24
A 500-page site contains about 12,000 links and accumulates broken references after every release, and you have one week to deliver a link-checking CI design; what would you implement?
reference-docsdesignlink-checking
25
One Git repository must publish 300 public pages and 90 authenticated partner pages, including NDA-only details; how would you deliver a content-partition and publishing architecture in three weeks?
gitarchitecturepartitioning
26
Fourteen OpenAPI specifications feed one developer portal, and you have 10 days to deliver required CI checks before the next release; which contract checks would you design?
openapidesign
27
The docs repository receives 40 pull requests a week, and reviewers cannot judge rendered changes, so you have one week to deliver a preview-deployment workflow; how would it work?
deploymentpull-requestspreview-deployments
28
Documentation checks run across 25 repositories, failures are ignored because ownership is unclear, and you have two weeks to deliver a CI ownership matrix and recovery runbook; what would you define?
documentationownershiprunbooks
29
Seventy documentation pull requests arrive each month and median review time is six days, and you have three weeks to deliver a review workflow with a two-day target; how would you redesign it?
code-reviewdocumentationpull-requests
30
Five product teams need to maintain 500 pages in one documentation monorepo, and you have four weeks to deliver the repository map and contribution model; how would you prevent both bottlenecks and uncontrolled structure?
documentationmonorepotracking
31
A product supports three major versions with 420 shared pages, and you have six weeks to deliver a versioned-documentation architecture; how would you avoid copying the entire site three times?
documentationarchitecture
32
Twenty-four API endpoints will be deprecated with a 90-day notice, and you have two weeks to deliver a deprecation communication plan and page pattern; what would you include?
endpointscommunicationcontent-deprecation
33
Two products release monthly from separate code branches, and you have 10 days to deliver a documentation release-branch policy; how would you keep fixes and features aligned with product versions?
documentation
34
An audit finds 90 repeated prerequisites and warnings across 300 pages, and you have four weeks to deliver a single-sourcing component inventory and migration plan; how would you choose what to reuse?
componentsmigrations
35
A 500-page portal is launching in six locales in eight weeks, and you must deliver the localization handoff package to two vendors; what would the package contain?
localizationprocurement
36
Six localization teams use conflicting translations for about 600 product terms, and you have 12 days to deliver a glossary in CSV and TBX form; how would you build and govern it?
localizationterminology-managementforms
37
The help center has 280 screenshots across three UI themes, and you have three weeks to deliver a screenshot source manifest and quarterly maintenance process; how would you control drift?
iacconcurrency
38
Command reference for 11 CLI packages is copied manually and drifts every release, and you have two weeks to deliver a source-maintenance pipeline design; how would you generate and review it?
ci-cdiacdesign
39
Localized builds occasionally fail hours before release, and you have one week to deliver a fallback policy for six locales without silently serving mixed languages; what would you define?
40
Documentation must cover four supported product versions in six locales, but the team can maintain only two translated versions, and you have six weeks to deliver a support matrix and publishing strategy; how would you resolve the combinations?
documentation
41
A portal receives two million page views per month but the team cannot connect traffic to user success, and you have 14 days to deliver a documentation analytics specification and dashboard wireframe; what would you measure?
design-collabdocumentation
42
Internal search handled 18,000 queries last month, and you have one week to deliver a ranked backlog from zero-result and reformulation data; how would you separate content gaps from search defects?
backlogqueriesdefects
43
Support closed 1,200 tickets last quarter and asks for better docs, and you have 10 days to deliver a ticket-to-content opportunity map; how would you avoid treating every ticket as a missing article?
44
Leadership wants a quarterly quality score for 400 pages, and you have two weeks to deliver a scorecard that cannot be gamed by publishing more content; how would you define it?
45
A 300-page platform spans eight technical domains and 32 potential SMEs, and you have three weeks to deliver an SME and reviewer operating model; how would you assign accountability without requiring everyone on every review?
46
Twenty features enter design each quarter but documentation learns about them after code freeze, and you have two weeks to deliver a documentation checkpoint for product design reviews; what would it require?
documentationdesign
47
Twelve delivery teams use different release checklists, and you have three weeks to deliver a documentation definition-of-done that works for minor fixes and major launches; how would you design it?
documentationdesign
48
Product wants 40 launch pages in two weeks, engineering can review only 12, and support wants troubleshooting first; how would you deliver a decision memo that resolves the stakeholder conflict?
troubleshootingcommunicationstakeholder-management
49
Six product domains share 300 pages and ownership changes every quarter, and you have 15 days to deliver a domain-maintenance registry and monthly cadence; how would you keep it current?
ownershipregistries
50
Two writers receive 75 documentation requests for the next quarter and have five days to deliver a domain roadmap; how would you prioritize the queue and show what will not be done?
roadmapdocumentationprioritization
51
A REST API removed the status_text field yesterday, but 36 of 80 documentation pages still use it. What would you do before the next release candidate in three days?
restdocumentation
52
Authentication changed from an api_key query parameter to a Bearer token, and 24 of 50 code examples still show the old method. How would you handle the transition?
authtokensqueries
53
The OpenAPI file marks customer.email as required, but three captured 200 responses omit it and the prose says it is optional. How would you resolve the drift?
openapiiac
54
Generated reference docs list four payment states, while the service returns a fifth state named review_pending in 7% of sandbox calls. What would you change?
reference-docs
55
The Python SDK quickstart calls client.jobs.create(), but version 3.2 exposes client.job_runs.create() and 900 users viewed the page this week. How would you respond?
quickstartspython
56
JavaScript SDK docs promise three automatic retries, but a network trace from SDK 6.4 shows no retry for HTTP 429. How would you investigate and document it?
promisesjavascripthttp
57
Webhook docs say delivery is attempted once, but delivery logs show up to six attempts over 24 hours and customers are creating duplicate orders. What would you publish?
webhooks
58
Four customers report that invoice.paid sometimes arrives before invoice.created, although the webhook guide implies chronological order. How would you fix the guidance?
webhooks
59
Version 1 endpoints shut down in 21 days, but the migration guide says customers have 45 days and still references two removed parameters. What would you do?
reference-docsmigration-guidesendpoints
60
Release notes claim CSV exports support 100,000 rows, but the shipped limit is 10,000 and 14 customers have already opened tickets. How would you correct the error?
release-notes
61
After enabling Docusaurus versioning, the v2.6 build fails with Duplicate routes found for /api/auth, while the current docs build locally. How would you diagnose it?
versioningdocusaurus
62
MkDocs builds without errors, but 17 new pages are absent from navigation and search. What would you check?
63
A GitHub Actions docs deployment began failing with Resource not accessible by integration after permissions were tightened. How would you isolate the cause?
ci-cddeploymentgithub-actions
64
Vale suddenly reports 1,240 errors on unchanged pages after a dependency update, blocking 19 pull requests. What would you do?
dependenciespull-requestsvale
65
A link checker reports 86 broken fragment links after headings were translated into Spanish, while the pages return HTTP 200. How would you diagnose and repair them?
http
66
The MDX build fails only on one of 320 pages with Unexpected character `{` at line 74. How would you find the actual defect?
defects
67
Algolia search returns deleted v1 pages above current v3 guides for the query create token. What would you inspect?
queriestokens
68
A redirect change sends /sdk/python to /python and then back to /sdk/python, causing a browser loop. How would you fix it safely?
python
69
The docs build passes on every laptop but fails in GitHub Actions with an unsupported Node API after a package upgrade. How would you make the diagnosis reproducible?
ci-cdgithub-actionsapi
70
An external-link check fails on 3 to 12 different URLs each night, but manual clicks work in the morning. How would you reduce noise without hiding real failures?
71
An accessibility audit finds that code-language tabs on 18 pages cannot be operated by keyboard, and screen readers do not announce the selected language. How would you fix them?
documentation-accessibilitykeyboard-accessibilitya11y
72
Searches for limits return valid pages from four products, but 61% of users reopen the results within 20 seconds and then add a product name. How would you diagnose the information architecture?
information-architecturearchitecture
73
A 2,800-word troubleshooting page covers six errors, but users apply its first clear-cache workaround to every case and 38% reopen their tickets. How would you improve the diagnostic path?
troubleshootingcaching
74
A generated 90-row API parameter table overflows at 375 pixels, and all five mobile test participants miss the required marker on nested fields. What would you change?
api
75
The same deployment guide serves administrators and application developers, and each group skips about half of its 18 steps. How would you redesign it?
deployment
76
Of 140 product pages, 52 have not been reviewed in 18 months and no engineering owner is recorded. How would you establish workable ownership?
ownership
77
A public preview of an unreleased feature appears in Google results because the preview host publishes the production sitemap and indexable canonical tags. How would you contain the incident?
incidentsindexes
78
A crawler finds 26 valid pages with no inbound internal links, including four pages that receive more than 500 monthly visits from Google. What would you do?
79
The docs use workspace, project, and account for the same object across 63 pages, causing users to search all three terms. How would you fix the information architecture?
information-architecturearchitecture
80
The API landing page offers 14 equal-weight links, and usability tests show 6 of 8 developers cannot identify where to make a first call. How would you restructure it?
api
81
A parameter changed from region_id to region in English, but 73 pages across five locales still contain one or both names. How would you coordinate the update?
82
A release adds 46 strings eight hours after the translation freeze for eight locales, and only English plus German can be ready on launch day. What would you decide?
83
The version 2.4 guide uses a 2.5-only command, while 31% of customers remain on 2.4. How would you correct the versioning incident?
incidentsversioning
84
After a UI redesign, 28 screenshots are stale in six locales, but the workflow itself has only four distinct states. How would you update them efficiently?
distinct
85
A PNG architecture diagram still shows two retired services, while the editable Mermaid source shows the current four-service flow. What would you do?
architecture
86
A shared prerequisites include appears on 22 pages, but changing its Docker requirement breaks instructions for three legacy versions. How would you preserve single sourcing?
docker
87
French, Japanese, and Portuguese guides each translated JSON property names inside code samples, and all 19 samples now fail. How would you prevent recurrence?
88
The version selector sends users from a Spanish 2.3 page to the English current page, losing both locale and task. How would you repair it?
89
A Japanese 4.0 guide loads a French 3.8 tutorial video because every locale and version reuses the same media key, and the captions describe removed controls. How would you repair the incident?
tutorialsincidents
90
Generated API reference says a timeout is 30 seconds, while a hand-written troubleshooting note says 60 seconds for versions 4.1 through 4.3. How would you create one reliable source?
reference-docstroubleshootingapi
91
One security SME has 18 documentation reviews pending, with a median wait of nine days, and a release is in six days. How would you unblock accurate work?
documentation
92
Two engineers give conflicting reviews: one says webhooks retry for 24 hours, the other says 72 hours. The release is tomorrow. What would you do?
resiliencewebhooks
93
At release readiness, 7 of 62 changed pages lack SME approval, two examples fail, and the product can still delay by 24 hours. What recommendation would you make?
health-checks
94
Page views rose 35% after a navigation change, but successful copy events on the quickstart fell from 62% to 44%. How would you interpret the result?
quickstarts
95
Support logged 126 OAuth tickets in one month, and 71 mention redirect_uri_mismatch across three frameworks. How would you turn the cluster into documentation work?
documentationoauth
96
Nine of 12 customers in interviews say the quickstart works, but they do not know whether their first import succeeded. What would you change?
quickstarts
97
A junior writer's first release note lists 23 internal ticket titles and no user impact, with publication due in two days. How would you mentor them through the revision?
mentoring
98
A junior writer has received the same feedback about untested code samples in three pull requests, and 5 of 14 samples fail. What would you do?
feedbackcode-reviewpull-requests
99
A feature launches in four days, but the UI labels changed twice and the product manager has not answered six documentation questions. How would you protect release readiness?
documentationhealth-checks
100
Six weeks after a docs update, billing setup tickets fell from 84 to 49 per month, but failed search queries for tax ID rose from 90 to 160. How would you decide the next change?
zero-to-onequeries

Perguntas

1You inherit 320 help-center pages with no owner data and have 10 working days before planning; how would you produce a content-audit workbook that supports deletion and rewrite decisions?

2Documentation is split across Confluence, GitHub, and a legacy CMS with about 480 pages, and owners discover stale material only through complaints; how would you deliver a lifecycle metadata and ownership schema in two weeks?

3A three-product SaaS suite has 260 pages grouped by internal team names, and product wants a taxonomy proposal in three weeks; how would you deliver a taxonomy decision record and controlled vocabulary?

4Users need six clicks to reach setup instructions in a 180-page portal, and you have 12 days to deliver a navigation-tree prototype; how would you redesign and validate it?

5The portal logged 1,500 unique queries last month, 22 percent returned no useful result, and you have five days to deliver a search-term and synonym map; what would you include?

6Five writers use inconsistent structures across 220 pages, and you have two weeks to deliver page-template specifications for tutorials, how-to guides, concepts, and references; how would you design them?

7A 400-page knowledge base must feed web, in-product help, and support macros, and engineering wants a content-model specification in three weeks; how would you model it?

8A 300-page site mixes explanation, procedures, lessons, and parameter tables on the same pages, and you have 15 working days to deliver a Diátaxis separation map; how would you decide what to split?

9Users complete setup but then lose their path between configuration, verification, and troubleshooting across a 500-page portal; how would you deliver a contextual-linking specification in two weeks?

10You must move 350 pages to a new information architecture in six weeks while preserving bookmarked URLs; how would you deliver the migration map and cutover plan?

11A new public API has 120 endpoints in an OpenAPI file and launches in four weeks; how would you deliver a reference section that is accurate but not merely generated output?

12Developers must choose between API keys and OAuth 2.0 for a partner API, and you have 10 days to deliver an authentication onboarding quickstart for a 15-minute first call; how would you structure it?

13The API supports JavaScript, Python, Java, and Go SDKs, and you have three weeks to deliver a runnable examples repository covering eight core workflows; how would you keep examples equivalent and maintainable?

14A platform emits 18 webhook event types and partners launch in two weeks; how would you deliver a webhook integration guide that covers signature verification, retries, and testing?

15Nine list endpoints use cursor pagination but clients keep treating cursors as page numbers, and you have five days to deliver a pagination pattern page; what would it contain?

16Six payment endpoints support idempotency keys and go live in one week; how would you deliver a client-facing idempotency guide that prevents duplicate charges?

17Support has identified 64 public API error codes with inconsistent messages, and you have two weeks to deliver an error catalog plus troubleshooting paths; how would you design it?

18Three products ship weekly and release notes arrive in different formats, and you have five days to deliver a changelog template and intake process; how would you standardize them?

19API v1 has 220 active customer integrations and will be retired in 90 days, and you have four weeks to deliver a v1-to-v2 migration guide; how would you structure and verify it?

20A comparison finds 35 differences between the production API and its OpenAPI description, and you have eight days to deliver a reconciliation report before SDK generation; how would you resolve source-of-truth conflicts?

Perguntas bloqueadas