Design a CI/CD step to auto-lint application content

What's really being asked

This question tests whether you can translate code-quality CI/CD practices into content-quality gates. Interviewers want to see that you treat UI copy, error messages, and documentation as production assets that require automated validation before deploy, not after. The signal is architectural thinking about rule design, tool selection, and failure handling.

The full answer

A strong answer walks through four decisions in order. First, define the rule taxonomy: terminology consistency like brand casing and banned words, structural checks like placeholder key completeness and i18n token coverage, and semantic checks like hardcoded user-facing strings in source files. Second, match the tool to the content format: use TextLint for Markdown documentation to enforce terminology casing as Netlify does, Vale or Write Good for prose style guides, and custom AST-based scripts or linters for JSX or template files to catch hardcoded strings. Third, integrate the step as a blocking CI job that runs in under a minute on changed files only, emitting annotated PR comments so writers and developers see exact line violations. Fourth, define severity levels: block deploy on missing i18n keys or broken placeholders, warn on style deviations, and provide a bypass mechanism for intentional exceptions.

The mistakes people make

Red flags include suggesting manual proofreading instead of automation, proposing a single generic spell-checker without domain-specific rules, or running linting post-deploy. Another weak pattern is failing to distinguish fatal errors from warnings, which trains teams to ignore the linter. Saying you would write everything from scratch without mentioning existing tools like Vale or TextLint also signals ignorance of the ecosystem.

What usually comes next

Expect the interviewer to ask how you would handle false positives from an aggressive terminology rule, how to enforce consistency across micro-frontends owned by different teams, or how you would version and share linting rules across repositories. They may also probe performance on large codebases and ask whether you would gate the main branch or every pull request.

A concrete example

Suppose your application uses React with Markdown-based help docs. Your pipeline step runs TextLint with a custom terminology rule that enforces Sign in over Log in and validates that all JSX Button components use an i18n key rather than a raw string. The job executes via GitHub Actions on pull requests, fails the build if TextLint finds casing violations or if a custom ESLint plugin detects untranslated strings, and posts a summary comment to the PR. If the linter detects violations, the pipeline stops so flawed content never reaches users, matching the Netlify approach of preventing bad docs from publishing.