What problem does it solve? Teams driving Azure through the az CLI repeatedly hit the same failures: long-lived service principal secrets committed to repos, scripts parsing -o table output, JMESPath predicates returning empty results from double-quoted strings, CI steps running against the wrong subscription, and sleep loops instead of proper waiters. This Skill encodes the full az contract — command grammar, identity-first authentication, output formats, JMESPath querying, configuration precedence, extensions, pagination, and CI hardening — so generated and reviewed az usage is reproducible, scoped, and secret-free. ## Core Features & Use Cases - Identity-first authentication playbook: managed identity for in-Azure compute, federated OIDC (azure/login@v2) for CI, certificate SPs as fallback, and least-privilege create-for-rbac scoping with --role and --scopes. - Output and query discipline: the seven output formats, client-side JMESPath --query rules (single-quoted strings, backtick literals, multiselect lists to pin tsv column order), and why -o table must never be parsed in scripts. - Operational mechanics: accounts/subscriptions/tenants, sovereign clouds (az cloud set before az login), two-tier config with AZURE_* env vars, extensions and dynamic install, --no-wait + az wait pollers, --ids @- batching, az rest, and the Azure MCP Server. - Read-only triage scripts: three bash tools (az-identity-check.sh, az-config-audit.sh, az-resource-inventory.sh) that snapshot identity, audit local config hygiene, and inventory resources without mutating anything. - Use Case: While reviewing a GitHub Actions workflow, flag a stored AZURE_CLIENT_SECRET, replace it with OIDC login, add --only-show-errors and telemetry-off settings, and rewrite a sleep 60 loop as az vm wait --created. ## Quick Start Ask the AI to review your shell script or GitHub Actions workflow that calls az commands and harden it for identity, output parsing, and CI use.