midnightntwrk/midnight-expert/devnet-health

Devnet Health

Status and health checks for the local Midnight devnet (node, indexer, proof server). Pure shell — requires only bash, docker, and curl.

The two checks answer different questions:

• Status — "Is the docker container present, and is it running?"
• Health — "Is the service responding to its health endpoint within 5 s?"

Container running ≠ service healthy: a container can be running while the service inside is still booting, crashing on a loop, or otherwise not serving requests. Run both checks for a complete picture.

Service Inventory

Scripts

Both scripts live in scripts/ next to this SKILL.md and accept an optional --json flag.

status.sh — container status

bash scripts/status.sh           # text: <service>\t<status>\t<containerName>
bash scripts/status.sh --json    # JSON: {"services":[...],"allRunning":bool}

Status values: running | stopped | not-found.

Exit 2 if Docker is missing or the daemon is not running (also reflected in JSON via the error field).

health.sh — HTTP probes

bash scripts/health.sh           # text: <service>\t<healthy|unhealthy>\t<ms>\t<httpCode>
bash scripts/health.sh --json    # JSON: {"node":{...},"indexer":{...},"proofServer":{...},"allHealthy":bool}

A service is healthy when its endpoint returns HTTP 2xx within the 5 s timeout.

Exit 2 if curl is not installed.

Usage Patterns

Quick eyeball check

bash scripts/status.sh
bash scripts/health.sh

"Is anything actually running?"

bash scripts/status.sh --json | jq -r '.allRunning'
bash scripts/health.sh --json | jq -r '.allHealthy'

Status overlaid with health (treat running-but-unhealthy as down)

status_json="$(bash scripts/status.sh --json)"
health_json="$(bash scripts/health.sh --json)"
 
# Example: node container is "running" but health endpoint failed → unhealthy
node_state="$(printf '%s' "$status_json" | jq -r '.services[] | select(.name=="node") | .status')"
node_healthy="$(printf '%s' "$health_json" | jq -r '.node.healthy')"
if [ "$node_state" = "running" ] && [ "$node_healthy" != "true" ]; then
  echo "node container is up but not serving"
fi

Per-service probe from a script (no jq)

while IFS=$'\t' read -r service status ms code; do
  case "$service" in
    node) [ "$status" = "healthy" ] && echo "node ok in ${ms}ms" ;;
  esac
done < <(bash scripts/health.sh)

Locating the Scripts at Runtime

Other skills, commands, and scripts that want to invoke these checks should resolve the absolute path via the plugin root. Use the midnight-plugin-utils:find-claude-plugin-root skill to materialize /tmp/cpr.py, then:

PLUGIN_ROOT="$(python3 /tmp/cpr.py midnight-tooling)"
SCRIPTS_ROOT="$PLUGIN_ROOT/skills/devnet-health/scripts"
 
bash "$SCRIPTS_ROOT/status.sh" --json
bash "$SCRIPTS_ROOT/health.sh" --json

This keeps the scripts callable regardless of where the plugin is installed (project vs user vs marketplace).

When To Use Which Skill

Compatibility Notes

• macOS / Linux: both scripts target POSIX bash 3.2+ and use only docker, curl, awk. No bash 4 features (associative arrays, mapfile).
• `jq` is optional: scripts do not depend on jq; they emit JSON directly with printf. Callers that want to parse JSON can use jq if available or fall back to the tab-separated text output.
• Timeouts: HTTP probes use a 5 s timeout. Status checks use Docker's default; on a healthy daemon they return immediately.