Build-time checks are designed to enforce security gates within the CI/CD pipeline. When critical vulnerabilities such as SQL injection and cross-site scripting (XSS) are detected during this stage, the correct and expected behavior is to fail the build. Stopping the build process prevents insecure code from progressing to later stages such as testing, deployment, or production.
Ignoring issues or merely sending alerts while continuing the pipeline undermines the purpose of shift-left security. Alerts to SIEM systems and issue trackers are typically supplementary actions, but the primary enforcement mechanism at build time is to block the pipeline when severity thresholds are exceeded. This approach reduces remediation costs, limits exposure, and ensures that only secure artifacts move forward in the DevSecOps lifecycle.

A pre-commit hook integrated with a secret-scanning tool (such as git-secrets, TruffleHog, or Gitleaks) runs locally on the developer's machine before a commit is finalized, scanning staged changes for patterns matching credentials like AWS access keys or API tokens, and blocking the commit if secrets are detected -- this stops leaks at the earliest possible point, before code is even pushed, exactly matching Ana's goal. Post-deployment log review happens far too late in the pipeline, after code has already been committed, pushed, built, and possibly deployed. A WAF rule update addresses runtime web application attack traffic, unrelated to preventing secret leakage in source control. Load testing evaluates application performance under load and has no bearing on credential leakage prevention. Since Ana needs to catch secrets before they're pushed to the remote repository, a pre-commit hook with secret scanning is correct.

Coverity is a well-known Static Application Security Testing (SAST) tool used to analyze source code for security vulnerabilities, coding errors, and quality issues. It integrates with CI/CD pipelines and supports enterprise-scale environments, including cloud-based development on platforms such as Azure. Accurics focuses on Infrastructure as Code security, Tenable.io is a vulnerability management platform for infrastructure and assets, and DevSkim is a lightweight code scanning extension rather than a full SAST platform. Selecting Coverity enables deep static analysis of application code during the Code stage, helping teams detect vulnerabilities early and reduce remediation costs.

Syft is used by Anchore to generate Software Bill of Materials (SBOMs) from container images and directories. By default, Syft may only analyze the squashed image view. Using the --scope all-layers flag instructs Syft to include software components fromall image layers, ensuring comprehensive visibility into dependencies introduced at every stage of image creation. The other options use invalid syntax or unsupported flags. Including all layers during SBOM generation improves vulnerability detection accuracy and supports compliance requirements, making it a critical practice during the Build and Test stage.

Data masking (or anonymization/pseudonymization) techniques replace or obscure sensitive fields such as names, national ID numbers, or medical record identifiers with realistic but non- identifying substitute values, while preserving data format, structure, and referential integrity needed for effective QA testing -- exactly satisfying Sara's dual requirement of privacy protection and testing realism. Fully replicating production data with no changes would expose real patient identities in a lower-security environment, violating privacy regulations like HIPAA or GDPR.
Disabling the staging environment entirely would prevent QA testing altogether, which is not a viable solution to the data privacy problem. Encrypting the entire database with a single shared key visible to all developers does not actually anonymize the data -- anyone with that key could still view real patient identities, and broad key visibility itself creates a serious exposure risk.
Since Sara needs privacy-preserving yet realistic test data, data masking/anonymization is correct.