tanstack-start-architecture

This skill is an aggressive, repository-local architecture enforcement policy checklist that itself contains no direct malicious payload or explicit network exfiltration. The primary security concerns are operational and supply-chain in nature: (1) mandatory use of an external persistent companion skill (/oh-my-claudecode:ralph) establishes a transitive trust/execution chain with unknown behavior and potential for data exfiltration or persistent unauthorized changes; (2) the agent is expected to read and potentially write widely across the repository (routes, features, database, config), increasing the privilege surface; (3) the rigid 'no exceptions' automation model risks autonomous modifications or denial-of-change without human review. Recommendations: do not auto-install or auto-invoke the companion skill without code review; restrict agent privileges to read-only unless explicit human approval is given for writes; log and review all outputs before sending them to external skills; and treat stylistic rules as configurable to avoid blocking legitimate work.