0.25.1 — get_health_summary destroyed filter + create_entity footgun doc

- get_health_summary: filter out state=destroyed entities (was noise
  from 20+ destroyed test LXCs, deprecated services, etc.)
- create_entity: document the monitoring footgun in the tool description
  (creating a type=check entity does NOT wire a check_def; the correct
  path is update_entity_attributes with monitoring + url attributes)

VERSION

@@ -1 +1 @@
-0.25.0
+0.25.1

internal/mcp/tools.go

@@ -113,7 +113,7 @@ func allTools(pool *db.Pool, agentID uuid.UUID) []toolReg {
 			query := `
 			SELECT e.slug, e.type, st.health, st.last_check_at
 			FROM entity_status st JOIN entities e ON e.id = st.entity_id
-			WHERE e.type <> 'check'`
+			WHERE e.type <> 'check' AND e.state <> 'destroyed'`
 			if healthStr != "" {
 				query += ` AND st.health = ANY(string_to_array($1, ','))`
 				return queryRows(ctx, pool, query, healthStr), nil
@@ -205,7 +205,7 @@ func allTools(pool *db.Pool, agentID uuid.UUID) []toolReg {
 			return upsertKnowledge(ctx, pool, args)
 		}},
 
-		{tool: &mcp.Tool{Name: "create_entity", Description: "Create a new entity in the knowledge graph — the creation half alongside update_entity_attributes (which only updates EXISTING entities). Use it when a task needs an entity that does not exist yet: a new check (check:<kind>:<target>:<n>), an ingress (ingress:<host>), a cert (cert:<host>), a service, a host/LXC/VM, etc. After inserting, it derives default checks from the entity type's monitoring spec (same as a seed ingest), so creating a checkable entity wires its monitoring in one call. Does NOT require approval (this updates the knowledge graph, not the live infrastructure). If the slug already exists it returns 'already exists' — then use update_entity_attributes to change it.",
+		{tool: &mcp.Tool{Name: "create_entity", Description: "Create a new entity in the knowledge graph. Use it when a task needs an entity that does not exist yet: a service, a host/LXC/VM, an ingress, a cert, etc. After inserting, it derives default checks from the entity type's monitoring spec, so creating a checkable entity wires its monitoring in one call. Does NOT require approval. If the slug already exists it returns 'already exists' — then use update_entity_attributes to change it. FOOTGUN: creating a type=check entity creates a bare entity row but does NOT wire a check_def — the scheduler will never probe it. To add monitoring, set `monitoring: [\"http\"]` + `url` on the target via update_entity_attributes.",
 			InputSchema: objSchema(
 				prop{"type", "string", "Entity type — must already exist in the ontology and not be abstract (e.g. service, lxc, host, vm, check, ingress, cert, dns)."},
 				prop{"name", "string", "Human-readable name (e.g. 'HAOS http service check')."},