From c35b10dea954b58530ff684e1d6dd4405f1340f3 Mon Sep 17 00:00:00 2001
From: Chris Burns <29541485+ChrisJBurns@users.noreply.github.com>
Date: Wed, 17 Jun 2026 16:21:53 +0100
Subject: [PATCH] fix(skill-build): use --host/--port for thv serve and correct
 health path

The skill-build action started the ToolHive daemon with
`thv serve --address`, a flag no released thv supports, so daemon
startup failed with "unknown flag: --address" (closes #16). Upstream
thv exposes `--host`/`--port` for `thv serve`.

Two further latent bugs from upstream drift are fixed alongside it:

- The readiness probe polled `/api/v1beta/health`, but the health
  route is mounted at top-level `/health`. The old path 404s, which
  would have timed out the 30s wait even after the flag fix.
- `/health` returns 503 when no container runtime is present, and
  `curl -sf` fails on 503. Skill build/push only package OCI
  artifacts and do not need a container runtime, so readiness now
  treats any HTTP response (including 503) as "daemon up". This also
  unblocks runners without Docker (relates to #3).

Adds a daily scheduled CI run so future breakage from upstream
ToolHive "latest" releases is caught proactively rather than at the
next push.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---
 .github/workflows/test.yml |  4 ++++
 skill-build/action.yml     | 11 +++++++----
 2 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 6da9797..e455b9a 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -6,6 +6,10 @@ on:
   pull_request:
     branches: [ main ]
   workflow_dispatch:
+  # Run daily so breakage from upstream ToolHive "latest" releases (e.g. renamed
+  # CLI flags or moved API routes) is caught proactively rather than at next push.
+  schedule:
+    - cron: '0 6 * * *'
 
 jobs:
   test-install-action:
diff --git a/skill-build/action.yml b/skill-build/action.yml
index 32daaef..2f89194 100644
--- a/skill-build/action.yml
+++ b/skill-build/action.yml
@@ -50,9 +50,12 @@ runs:
       id: daemon
       shell: bash
       run: |
-        # Check if a daemon is already reachable (e.g., started by run-mcp-server)
+        # Check if a daemon is already reachable (e.g., started by run-mcp-server).
+        # Any HTTP response (even 503 when no container runtime is present) means the
+        # API server is up. Skill build/push package OCI artifacts and do not require
+        # a container runtime, so we must not gate readiness on the runtime's health.
         EXISTING_URL="${TOOLHIVE_API_URL:-http://127.0.0.1:8080}"
-        if curl -sf "${EXISTING_URL}/api/v1beta/health" > /dev/null 2>&1; then
+        if [ "$(curl -s -o /dev/null -w '%{http_code}' "${EXISTING_URL}/health" 2>/dev/null)" != "000" ]; then
           echo "ToolHive daemon already running at $EXISTING_URL"
           PORT=$(echo "$EXISTING_URL" | sed -E 's|.*://[^:]+:([0-9]+).*|\1|')
           echo "port=$PORT" >> $GITHUB_OUTPUT
@@ -67,7 +70,7 @@ runs:
         }
 
         echo "Starting thv serve on port $PORT..."
-        thv serve --address "127.0.0.1:${PORT}" &
+        thv serve --host "127.0.0.1" --port "${PORT}" &
         DAEMON_PID=$!
         echo "pid=$DAEMON_PID" >> $GITHUB_OUTPUT
         echo "port=$PORT" >> $GITHUB_OUTPUT
@@ -82,7 +85,7 @@ runs:
             wait "$DAEMON_PID" 2>/dev/null
             exit 1
           fi
-          if curl -sf "http://127.0.0.1:${PORT}/api/v1beta/health" > /dev/null 2>&1; then
+          if [ "$(curl -s -o /dev/null -w '%{http_code}' "http://127.0.0.1:${PORT}/health" 2>/dev/null)" != "000" ]; then
             echo "ToolHive daemon is ready (port $PORT)"
             break
           fi