Verifier-First Evaluation of Agentic LLMs for Infrastructure-as-Code Generation

Collection of recent arXiv papers describing advances and evaluations in LLM-driven code and infrastructure generation, test synthesis, deployment validation, and AI-assisted software maintenance with clear security implications. A verifier-first study of Terraform generation shows LLMs can produce a high rate of invalid or policy-violating IaC (dominant failure modes: terraform validate failures and OPA/Rego information-gap failures) but that retrieval, verifier feedback, and prompt/teacher techniques substantially raise pass rates; showing that exposing policy text and integrating verifiers/