pilotswarm-aks-deploy

# PilotSwarm AKS deployment Read `.github/DEPLOYMENT.md` first to select the target and execution path. For this repository's managed Azure environment, use the documented Actions from `main`; local deployment/reset commands are not an alternative. Keep real identifiers and credentials in ignored config and protected secrets. For a fresh machine, local provider tests or a fork's CI installation, follow `docs/developer/contributing/local-ci-and-tests.md` and `.github/CI.md`. For releases use the `pilotswarm-release` skill: tarballs in GitHub Releases, complete PostgreSQL baseline plus additive HDB gate, then Azure deployment. Ordinary merges do not deploy. Release publication requires a publication request. Local `deploy:new-env` scaffolding is allowed; inspect existing ignored stamp config before creating or replacing files. Resolve the intended auth posture before creating an Entra app or assigning access. Reusing an existing app does not require creating another one. Match the operator's tenant requirements. Only for an explicitly selected operator-managed environment, consult [the detailed operations reference](references/legacy-aks.md). Do not assume bare “the cluster” means a legacy bash target. Keep the selected controller, resource group and namespace consistent. Deployment never implies permission to wipe data. Preserve already-granted authorization; request only missing target information or authorization needed for a new operation. Verify workflow completion, deployed source/images, readiness and the public portal assets before declaring success. Keep output free of private endpoints and credentials. Read-only local inspection is permitted.