fix(node): mirror official Grok hooks

integrations/README.md

@@ -4,7 +4,7 @@
 
 The provider-level contract represents the default stable runtime. `contract_variants` records opt-in or transitional runtimes with their activation command and release channel. Kiro CLI uses this field for its v2 embedded hooks and v3 standalone hooks.
 
-Grok Build adapters write native user hook files under `$GROK_HOME/hooks` or `~/.grok/hooks`. Project `.grok/hooks` files require `/hooks-trust` or `--trust`; Claude Code and Cursor hook files are read-only compatibility sources for Elydora.
+Grok Build adapters write exact `PreToolUse`, `PostToolUse`, and `PostToolUseFailure` groups under `$GROK_HOME/hooks` or `~/.grok/hooks`. The native camelCase payload includes `toolUseId`, `toolInputTruncated`, and `toolResult`; explicit denials use JSON stdout plus exit code `2`, while hook crashes and timeouts fail open. Project `.grok/hooks` files require `/hooks-trust` or `--trust`; `hooks-paths`, plugin, Claude Code, and Cursor sources remain read-only for Elydora.
 
 GitHub Copilot command `preToolUse` hooks deny on crashes and every non-zero exit. Command timeouts continue through the normal permission flow, so `timeout_failure_mode` records that explicit exception when it differs from `failure_mode`.
 

integrations/catalog.json

@@ -214,10 +214,24 @@
       "config_paths": [
         "$GROK_HOME/hooks/*.json",
         "~/.grok/hooks/*.json",
-        ".grok/hooks/*.json"
+        "~/.grok/hooks-paths",
+        ".grok/hooks/*.json",
+        ".claude/settings.json",
+        ".cursor/hooks.json"
       ],
-      "events": { "before_tool": "PreToolUse", "after_tool": "PostToolUse" },
-      "event_fields": { "name": "toolName", "input": "toolInput", "session": "sessionId" },
+      "events": {
+        "before_tool": "PreToolUse",
+        "after_tool": "PostToolUse",
+        "after_tool_failure": "PostToolUseFailure"
+      },
+      "event_fields": {
+        "name": "toolName",
+        "input": "toolInput",
+        "session": "sessionId",
+        "call_id": "toolUseId",
+        "output": "toolResult",
+        "error": "toolResult"
+      },
       "blocking": { "mechanism": "exit_code_2", "failure_mode": "fail_open" },
       "adapters": { "node": true, "python": true, "go": true },
       "delivery_state": "available",