Cloud Resilience and Dependency Auditor — Find Your Single Points of Failure Before the Next Outage Does

About This Skill

Every outage post-mortem recommends the same first step, and almost nobody has done it: map every critical service, API, and route your uptime depends on — not just the data plane, but the control plane. The dependency tree is no longer obvious from the invoice, and "we're multi-region" is not the same as "we're resilient," because what takes you down is usually a shared service — DNS, identity, a control plane — that is itself a single point of failure across all your regions. Cloud Resilience and Dependency Auditor does the audit. Describe your stack — providers and regions, and the services you lean on for DNS, identity, CDN, queues, data, secrets, observability, and CI/CD — and it maps the real dependency tree including the control-plane and hidden shared-service couplings people miss, ranks the single points of failure by blast radius, and checks the classic traps: monitoring and recovery tooling that share fate with the environment that's failing (so you go blind exactly when you need clarity), auth as a universal gatekeeper, the default-region hub, and vendors whose dependency tree you inherit. It returns a dependency map, a ranked single-point-of-failure register noting whether you could even recover while each is down, a resilience backlog prioritized by blast radius times likelihood with the specific fix and rough effort for each, and an accepted-risk register for what to consciously live with. The download includes three reference files: the dependency-inventory worksheet, a single-point-of-failure pattern guide, and a worked sample audit. It audits from what you describe — it doesn't scan your infrastructure or test failover, and a plan you haven't rehearsed is a hypothesis. Works with Claude Code, Cursor, Codex CLI, Gemini CLI, and any SKILL.md agent.