refactor: remove HTML markers from voting participant agents

- Remove HTML comment markers from AI_Moderator and AI_Designer
- These are voting participants, not background tools
- Should comment naturally like humans in discussions
- Occasional duplicate comments are acceptable and enrich discussion
- Background agents (Visualizer, Researcher) still keep markers
- Updated test to check for 'Name: AI_Moderator' instead of markers
- All 41 tests passing

agents/designer.py

@@ -54,9 +54,6 @@ def load_provider_client(repo_root: Path):
     return ProviderClient
 
 
-BLOCK_START = "<!-- AUTO:DESIGNER START -->"
-BLOCK_END = "<!-- AUTO:DESIGNER END -->"
-
 PRIMER = dedent(
     """
     You are AI_Designer, a senior UX designer collaborating with the CascadingDev
@@ -95,12 +92,11 @@ def scrub_comment(text: str) -> str:
 
 
 def append_designer_comment(context, comment: str, vote: str) -> None:
-    lines = [BLOCK_START, "Name: AI_Designer"]
+    lines = ["---", "", "Name: AI_Designer"]
     body = scrub_comment(comment)
     if body:
         lines.append(body)
     lines.append(f"VOTE: {vote}")
-    lines.append(BLOCK_END)
     lines.append("")
     context.append_block("\n".join(lines))
 

agents/moderator.py

@@ -47,9 +47,6 @@ def load_provider_client(repo_root: Path):
 
 # --- Agent Configuration ---
 
-BLOCK_START = "<!-- AUTO:MODERATOR START -->"
-BLOCK_END = "<!-- AUTO:MODERATOR END -->"
-
 PRIMER = """
 You are AI_Moderator, the project steward for CascadingDev discussions.
 
@@ -81,12 +78,11 @@ def scrub_comment(text: str) -> str:
     return "\n".join(cleaned).strip()
 
 def append_moderator_comment(context, comment: str, vote: str) -> None:
-    lines = [BLOCK_START, "Name: AI_Moderator"]
+    lines = ["---", "", "Name: AI_Moderator"]
     comment = scrub_comment(comment)
     if comment:
         lines.append(comment)
     lines.append(f"VOTE: {vote}")
-    lines.append(BLOCK_END)
     lines.append("")
     context.append_block("\n".join(lines))
 

tests/test_workflow.py

@@ -898,14 +898,14 @@ class ProviderClient:
 
     runner.process(repo, rules, model)
     first_pass = discussion.read_text(encoding="utf-8")
-    assert first_pass.count("<!-- AUTO:MODERATOR START -->") == 1
+    assert first_pass.count("Name: AI_Moderator") == 1
 
     # Stage and run again; moderator should contribute a fresh comment.
     run_git(repo, "add", ".")
     runner.process(repo, rules, model)
     second_pass = discussion.read_text(encoding="utf-8")
 
-    assert second_pass.count("<!-- AUTO:MODERATOR START -->") == 2
+    assert second_pass.count("Name: AI_Moderator") == 2
 
 
 def test_visualizer_generates_diagram(temp_repo):