fix(delegate): flatten content blocks in live overlay tail + AUTHOR_MAP

Follow-up on the cherry-picked content-block fix. _extract_output_tail
(the live subagent overlay) still used crude str(content), which renders
a "[{'type': 'text'...}]" blob and — worse — mislabels a block-wrapped
"Error: ..." result as is_error=False. Route it through the same
_stringify_tool_content helper so error detection and previews work at
both consumer sites.

- delegate_tool.py: _extract_output_tail uses _stringify_tool_content
- tests: add _extract_output_tail content-block test (error detection +
  clean preview)
- release.py: AUTHOR_MAP entry for randomsnowflake (CI gate)

scripts/release.py

@@ -45,6 +45,7 @@ ACP_REGISTRY_MANIFEST = REPO_ROOT / "acp_registry" / "agent.json"
 
 # Auto-extracted from noreply emails + manual overrides
 AUTHOR_MAP = {
+    "al@randomsnowflake.me": "randomsnowflake",
     "834740219@qq.com": "ViewWay",
     "harjoth.khara@gmail.com": "harjothkhara",
     "129007007+HeLLGURD@users.noreply.github.com": "HeLLGURD",

tests/tools/test_delegate.py

@@ -28,6 +28,7 @@ from tools.delegate_tool import (
     _build_child_agent,
     _build_child_progress_callback,
     _build_child_system_prompt,
+    _extract_output_tail,
     _strip_blocked_tools,
     _resolve_child_credential_pool,
     _resolve_delegation_credentials,
@@ -585,6 +586,40 @@ class TestDelegateObservability(unittest.TestCase):
             self.assertEqual(trace[0]["status"], "ok")
             self.assertGreater(trace[0]["result_bytes"], 0)
 
+    def test_output_tail_flattens_list_content_blocks(self):
+        """_extract_output_tail (live overlay) must flatten content-block lists
+        so error markers buried inside blocks are detected and previews are
+        real text, not a "[{'type': 'text'...}]" repr blob."""
+        result = {
+            "messages": [
+                {"role": "assistant", "tool_calls": [
+                    {"id": "t1", "function": {"name": "terminal", "arguments": "{}"}}
+                ]},
+                {"role": "tool", "tool_call_id": "t1", "content": [
+                    {"type": "text", "text": "Error: command not found"},
+                ]},
+                {"role": "assistant", "tool_calls": [
+                    {"id": "t2", "function": {"name": "vision", "arguments": "{}"}}
+                ]},
+                {"role": "tool", "tool_call_id": "t2", "content": [
+                    {"type": "text", "text": "all good"},
+                    {"type": "image_url", "image_url": {"url": "data:x"}},
+                ]},
+            ]
+        }
+        tail = _extract_output_tail(result, max_entries=8, max_chars=600)
+        by_tool = {t["tool"]: t for t in tail}
+
+        # Block-wrapped error is correctly flagged (crude str() would miss it).
+        self.assertTrue(by_tool["terminal"]["is_error"])
+        self.assertEqual(by_tool["terminal"]["preview"], "Error: command not found")
+        # Non-error multimodal result is not flagged, and the text is readable.
+        self.assertFalse(by_tool["vision"]["is_error"])
+        self.assertIn("all good", by_tool["vision"]["preview"])
+        # No raw content-block repr leaked into any preview.
+        for entry in tail:
+            self.assertNotIn("'type'", entry["preview"])
+
     def test_tool_trace_detects_error(self):
         """Tool results containing 'error' should be marked as error status."""
         parent = _make_mock_parent(depth=0)

tools/delegate_tool.py

@@ -261,9 +261,11 @@ def _extract_output_tail(
             break
         if not isinstance(msg, dict) or msg.get("role") != "tool":
             continue
-        content = msg.get("content") or ""
-        if not isinstance(content, str):
-            content = str(content)
+        # Flatten content-block lists/dicts to text so the overlay shows real
+        # output (not a "[{'type': 'text'...}]" blob) and error detection can
+        # see markers buried inside content blocks. Crude str() here would
+        # mislabel a block-wrapped "Error: ..." result as is_error=False.
+        content = _stringify_tool_content(msg.get("content") or "")
         is_error = _looks_like_error_output(content)
         tool_name = pending_call_by_id.get(msg.get("tool_call_id") or "", "tool")
         # Preserve line structure so the overlay's wrapped scroll region can