| name | health |
| description | Manage the hal-health sidecar container that serves live ecosystem state snapshots to HAL Plus and other consumers. Use when the user asks to refresh the status snapshot, recreate the health container, or troubleshoot HAL Plus product state. |
HAL Health Skill
The hal health namespace manages the hal-health sidecar container. It builds a frozen snapshot of the running HAL ecosystem and serves it at http://hal-health:9001/api/status on hal-net so HAL Plus and AI clients can query product state without touching the container engine directly.
Primary Commands
hal health create # create (or recreate) the hal-health container with a fresh snapshot
hal health update # refresh the snapshot — use after ecosystem changes outside normal lifecycle
hal health delete # remove the hal-health container
When To Use
hal health create: first-time setup or after hal health delete.hal health update: manual escape hatch after a product was added/removed outside the normal hal <product> create/delete flow (e.g. manual container changes).hal health delete: clean up the sidecar without affecting any product containers.
Architecture Notes
- The
hal-health container reuses ghcr.io/hashimiche/hal-mcp:latest with --entrypoint /usr/local/bin/hal and health _serve as args.
- The snapshot shape:
{ timestamp, engine, products: [{ product, state, health, reason, endpoint, containers, features: [...] }] }.
- The container is a no-op host — it reads the
HAL_HEALTH_DATA env var injected at start time and never touches the container engine socket.
global.RefreshHalHealth is called automatically after all product create/delete commands — manual hal health update is only needed for out-of-band changes.hal health is a no-op if hal-net does not exist or ghcr.io/hashimiche/hal-mcp:latest is not present locally.
Lab Assumptions
- HAL Plus must be running on
hal-net for hal-health to be reachable by HAL Plus.
- HAL Plus falls back to direct product endpoint probing if
hal-health is unavailable.
