diff --git a/graph/solutions/ct-302-health-checkpy-http-probes-docker-health-detection-wa-663f99.md b/graph/solutions/ct-302-health-checkpy-http-probes-docker-health-detection-wa-663f99.md
new file mode 100644
index 00000000000..b96e2bd5ad9
--- /dev/null
+++ b/graph/solutions/ct-302-health-checkpy-http-probes-docker-health-detection-wa-663f99.md
@@ -0,0 +1,51 @@
+---
+id: 663f99a9-e31b-4a02-a678-9837345aebbb
+type: solution
+title: "CT 302 health_check.py: HTTP probes, Docker health detection, warning accumulator"
+tags: [claude-runner, monitoring, health-check, homelab, docker, http-probes, discord]
+importance: 0.7
+confidence: 0.8
+created: "2026-02-20T06:35:19.624994+00:00"
+updated: "2026-02-20T06:35:19.624994+00:00"
+---
+
+# CT 302 Health Check Enhancements
+
+## Context
+Enhanced `health_check.py` on CT 302 (claude-runner, 10.10.0.148) with three new monitoring capabilities. Repo: `cal/claude-runner-monitoring` on Gitea.
+
+## Capabilities Added
+
+### 1. HTTP Health Probes
+8 endpoints across 3 servers, run directly from CT 302 via `requests` library (no SSH hop):
+
+**arr-stack:**
+- Sonarr: `:8989/ping` → `{"status":"OK"}`
+- Radarr: `:7878/ping` → `{"status":"OK"}`
+- Readarr: `:8787/ping` → `{"status":"OK"}`
+- Lidarr: `:8686/ping` → `{"status":"OK"}`
+- Jellyseerr: `:5055/api/v1/status` → 200
+- SABnzbd: `:8080/api?mode=version` → 200
+
+**Gitea:** `:3000/api/v1/version` → 200
+
+**Uptime Kuma:** `:3001/api/entry-page` → 200
+
+Note: Prowlarr is NOT deployed on arr-stack (omitted from probes).
+
+Catches: status code mismatches, timeouts, connection errors.
+
+### 2. Docker Health/Restart Detection
+Enhanced `docker ps` format string to include `{{.Status}}`. Detects:
+- `(unhealthy)` containers — auto-remediable
+- Uptime < 5 min — indicates restart loop, NOT auto-remediable
+
+Helper function `_uptime_seconds()` parses Docker status strings (e.g. "Up 3 minutes").
+
+### 3. Warning Accumulator
+Persistent `.warning_state.json` tracks consecutive warning counts per server. After 6 consecutive checks (30 min at 5-min interval), sends Discord digest via `send_warning_digest()` in `notifier.py`. Resets on clean run or after digest sent.
+
+## Deployment
+- Committed as `2b6e59a` on CT 302 (local only — Gitea token expired, see related memory)
+- Venv at `/root/.claude/skills/server-diagnostics/.venv/` with `requests`, `pyyaml`
+- Files deployed via `scp` to `/root/.claude/skills/server-diagnostics/`