What problem does it solve? Terragrunt live repos and catalogs hide orchestration-layer bugs — dependency cycles, unsafe mock outputs, broken state isolation, and tangled include hierarchies — that plain Terraform linters never see. This skill audits that orchestration layer with multiple independent LLM reviewers and merges their findings into one cross-validated report. ## Core Features & Use Cases - Multi-agent review: Dispatches one or more configured LLMs (claude, codex, gemini) as independent subagents, then merges and cross-validates their findings with severity filtering. - Orchestration-focused rubric: Checks dependency/DAG correctness, mock-output schemas, remote_state and generate blocks, state key isolation, include hierarchy, git source refspecs, and the DRY values pattern. - Consent-gated safety: Dry-run preview, explicit consent before sending HCL to provider CLIs, and fail-closed flags for scanner execution and PR posting. - Use Case: Point it at a live/prod Terragrunt tree before a rollout to catch a dependency cycle and a mock-output mismatch across environments, with findings confirmed by two independent models. ## Quick Start Ask the assistant to review the Terragrunt live directory for dependency and state issues, starting with a dry-run preview before approving the multi-agent dispatch.
