deployment-verification

Release gate that catches "green build, healthy container, broken page" deploys: CSS audit, computed-style assertions, screenshot diff, and Docker candidate promotion with verified rollback.

Mechanical enforcement tools to prevent AI agents from bypassing established project standards.

Herramientas de QA web, 8 capacidades CSS/WebGL autocontenidas y plantillas — todo probado en producción | Faro Digital, Barcelona

GitHub Action: scan a deployed URL with BreachProbe and fail the build on an exposed database or an open write path.

Claude Code skill: exploratory smoke-sweep + deploy verification for running/deployed web apps. Catches bugs that pass typecheck and return 200 but are wrong (dead DB columns, silent query failures, fake stats).

The Trust Stack for Agent Apps — independently installable Python libraries that make AI agents reliable, auditable, and accountable. Because agent failures are trust failures.

验证 agent 的完成声明：报 PASS 的检查器必须先自证能报 FAIL。假成功 / 静默失败判据库 56 条，每条自带反例。Verify-before-claim criteria library against silent completion claims.

Audit redirects, caching, compression, security headers, TLS, and delivery performance.