diff --git a/CLAUDE.md b/CLAUDE.md
index 441f7332e5..302eaa9188 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -234,3 +234,12 @@ Check in pyproject.toml and NEVER use a marker/tag that doesn't exist there. Ask
   - `runbooks: {}` - No runbooks available (empty catalog)
   - `runbooks: {catalog: [...]}` - Custom runbook catalog with entries pointing to .md files in the same directory
   - If `runbooks` field is not specified, default system runbooks are used
+
+## Documentation Lookup
+
+When asked about content from the HolmesGPT documentation website (https://robusta-dev.github.io/holmesgpt/), look in the local `docs/` directory:
+- Python SDK examples: `docs/installation/python-installation.md`
+- CLI installation: `docs/installation/cli-installation.md`
+- Kubernetes deployment: `docs/installation/kubernetes-installation.md`
+- Toolset documentation: `docs/data-sources/builtin-toolsets/`
+- API reference: `docs/reference/`
diff --git a/docs/installation/python-installation.md b/docs/installation/python-installation.md
index 12a3368322..e3ada2e0c1 100644
--- a/docs/installation/python-installation.md
+++ b/docs/installation/python-installation.md
@@ -18,7 +18,8 @@ pip install "https://github.com/robusta-dev/holmesgpt/archive/refs/heads/master.
 ```python
 import os
 from holmes.config import Config
-from holmes.plugins.prompts import load_and_render_prompt
+from holmes.core.prompt import build_initial_ask_messages
+from rich.console import Console
 
 print("🚀 Initializing HolmesGPT...")
 
@@ -31,22 +32,28 @@ config = Config(
 )
 print(f"✅ Configuration created with model: {config.model}")
 
-# Create AI instance
+# Create AI instance and console
 print("Creating AI instance...")
 ai = config.create_console_toolcalling_llm()
+console = Console()
 print("✅ AI instance ready")
 
 # Ask a question
-print("Loading system prompt...")
-system_prompt = load_and_render_prompt(
-    "builtin://generic_ask.jinja2",
-    {"toolsets": ai.tool_executor.toolsets}
+question = "what pods are failing in production?"
+print(f"\n🔍 Asking: '{question}'")
+
+# Build initial messages with system prompt
+messages = build_initial_ask_messages(
+    console=console,
+    initial_user_prompt=question,
+    file_paths=None,
+    tool_executor=ai.tool_executor,
+    runbooks=config.get_runbook_catalog(),
+    system_prompt_additions=None
 )
-print("✅ System prompt loaded")
 
-print("\n🔍 Asking: 'what pods are failing in production?'")
 print("Holmes is thinking...")
-response = ai.prompt_call(system_prompt, "what pods are failing in production?")
+response = ai.call(messages)
 print(f"Holmes: {response.result}")
 ```
 
@@ -62,7 +69,8 @@ Complete example of using HolmesGPT Python SDK with progress tracking
 
 import os
 from holmes.config import Config
-from holmes.plugins.prompts import load_and_render_prompt
+from holmes.core.prompt import build_initial_ask_messages
+from rich.console import Console
 
 def main():
     print("🚀 Starting HolmesGPT Python SDK Example")
@@ -81,8 +89,9 @@ def main():
     print(f"✅ Configuration created with model: {config.model}")
 
     print("\nStep 2: Creating AI instance...")
-    # Create AI instance
+    # Create AI instance and console
     ai = config.create_console_toolcalling_llm()
+    console = Console()
     print("✅ AI instance created successfully")
 
     print("\nStep 3: Listing available toolsets...")
@@ -99,16 +108,7 @@ def main():
     for tool in sorted(available_tools):
         print(f"   • {tool}")
 
-    print("\nStep 5: Loading system prompt...")
-    # Load system prompt
-    system_prompt = load_and_render_prompt(
-        "builtin://generic_ask.jinja2",
-        {"toolsets": ai.tool_executor.toolsets}
-    )
-    print("✅ System prompt loaded successfully")
-    print(f"Prompt length: {len(system_prompt)} characters")
-
-    print("\nStep 6: Asking questions...")
+    print("\nStep 5: Asking questions...")
     # Ask questions
     questions = [
         "what pods are failing in production?",
@@ -122,7 +122,18 @@ def main():
 
         try:
             print("Holmes is thinking...")
-            response = ai.prompt_call(system_prompt, question)
+
+            # Build initial messages
+            messages = build_initial_ask_messages(
+                console=console,
+                initial_user_prompt=question,
+                file_paths=None,
+                tool_executor=ai.tool_executor,
+                runbooks=config.get_runbook_catalog(),
+                system_prompt_additions=None
+            )
+
+            response = ai.call(messages)
             print(f"Holmes: {response.result}")
 
             # Show tools that were used
@@ -200,20 +211,19 @@ def main():
     ai = config.create_console_toolcalling_llm()
     console = Console()
 
-    # Load system prompt
-    system_prompt = load_and_render_prompt(
-        "builtin://generic_ask.jinja2",
-        {"toolsets": ai.tool_executor.toolsets}
-    )
-
     # First question
     print("\n🔍 First Question:")
     first_question = "what pods are failing in my cluster?"
     print(f"User: {first_question}")
 
-    # Build initial messages (system + first user message)
+    # Build initial messages (includes system prompt + first user message)
     messages = build_initial_ask_messages(
-        console, system_prompt, first_question, None
+        console=console,
+        initial_user_prompt=first_question,
+        file_paths=None,
+        tool_executor=ai.tool_executor,
+        runbooks=config.get_runbook_catalog(),
+        system_prompt_additions=None
     )
 
     # Call AI with initial messages
diff --git a/holmes/core/prompt.py b/holmes/core/prompt.py
index c9bffd0bf4..e2e8ef4fd0 100644
--- a/holmes/core/prompt.py
+++ b/holmes/core/prompt.py
@@ -1,11 +1,13 @@
 from rich.console import Console
-from typing import Optional, List, Dict
+from typing import Optional, List, Dict, Any, Union
 from pathlib import Path
+from holmes.plugins.prompts import load_and_render_prompt
+from holmes.plugins.runbooks import RunbookCatalog
 
 
 def append_file_to_user_prompt(user_prompt: str, file_path: Path) -> str:
     with file_path.open("r") as f:
-        user_prompt += f"\n\n<attached-file path='{file_path.absolute()}>'\n{f.read()}\n</attached-file>"
+        user_prompt += f"\n\n<attached-file path='{file_path.absolute()}'>\n{f.read()}\n</attached-file>"
 
     return user_prompt
 
@@ -25,11 +27,34 @@ def append_all_files_to_user_prompt(
 
 def build_initial_ask_messages(
     console: Console,
-    system_prompt_rendered: str,
     initial_user_prompt: str,
     file_paths: Optional[List[Path]],
+    tool_executor: Any,  # ToolExecutor type
+    runbooks: Union[RunbookCatalog, Dict, None] = None,
+    system_prompt_additions: Optional[str] = None,
 ) -> List[Dict]:
-    """Build the initial messages for the AI call."""
+    """Build the initial messages for the AI call.
+
+    Args:
+        console: Rich console for output
+        initial_user_prompt: The user's prompt
+        file_paths: Optional list of files to include
+        tool_executor: The tool executor with available toolsets
+        runbooks: Optional runbook catalog
+        system_prompt_additions: Optional additional system prompt content
+    """
+    # Load and render system prompt internally
+    system_prompt_template = "builtin://generic_ask.jinja2"
+    template_context = {
+        "toolsets": tool_executor.toolsets,
+        "runbooks": runbooks or {},
+        "system_prompt_additions": system_prompt_additions or "",
+    }
+    system_prompt_rendered = load_and_render_prompt(
+        system_prompt_template, template_context
+    )
+
+    # Append files to user prompt
     user_prompt_with_files = append_all_files_to_user_prompt(
         console, initial_user_prompt, file_paths
     )
diff --git a/holmes/interactive.py b/holmes/interactive.py
index 522f33aee8..4171e67092 100644
--- a/holmes/interactive.py
+++ b/holmes/interactive.py
@@ -761,12 +761,13 @@ def display_recent_tool_outputs(
 def run_interactive_loop(
     ai: ToolCallingLLM,
     console: Console,
-    system_prompt_rendered: str,
     initial_user_input: Optional[str],
     include_files: Optional[List[Path]],
     post_processing_prompt: Optional[str],
     show_tool_output: bool,
     tracer=None,
+    runbooks=None,
+    system_prompt_additions: Optional[str] = None,
 ) -> None:
     # Initialize tracer - use DummyTracer if no tracer provided
     if tracer is None:
@@ -965,7 +966,12 @@ def get_bottom_toolbar():
 
             if messages is None:
                 messages = build_initial_ask_messages(
-                    console, system_prompt_rendered, user_input, include_files
+                    console,
+                    user_input,
+                    include_files,
+                    ai.tool_executor,
+                    runbooks,
+                    system_prompt_additions,
                 )
             else:
                 messages.append({"role": "user", "content": user_input})
diff --git a/holmes/main.py b/holmes/main.py
index d510c7e779..9f8adecf25 100644
--- a/holmes/main.py
+++ b/holmes/main.py
@@ -180,10 +180,6 @@ def ask(
     destination: Optional[DestinationType] = opt_destination,
     slack_token: Optional[str] = opt_slack_token,
     slack_channel: Optional[str] = opt_slack_channel,
-    # advanced options for this command
-    system_prompt: Optional[str] = typer.Option(
-        "builtin://generic_ask.jinja2", help=system_prompt_help
-    ),
     show_tool_output: bool = typer.Option(
         False,
         "--show-tool-output",
@@ -214,6 +210,11 @@ def ask(
         "--trace",
         help="Enable tracing to the specified provider (e.g., 'braintrust')",
     ),
+    system_prompt_additions: Optional[str] = typer.Option(
+        None,
+        "--system-prompt-additions",
+        help="Additional content to append to the system prompt",
+    ),
 ):
     """
     Ask any question and answer using available tools
@@ -252,12 +253,6 @@ def ask(
         refresh_toolsets=refresh_toolsets,  # flag to refresh the toolset status
         tracer=tracer,
     )
-    template_context = {
-        "toolsets": ai.tool_executor.toolsets,
-        "runbooks": config.get_runbook_catalog(),
-    }
-
-    system_prompt_rendered = load_and_render_prompt(system_prompt, template_context)  # type: ignore
 
     if prompt_file and prompt:
         raise typer.BadParameter(
@@ -292,20 +287,23 @@ def ask(
         run_interactive_loop(
             ai,
             console,
-            system_prompt_rendered,
             prompt,
             include_file,
             post_processing_prompt,
             show_tool_output,
             tracer,
+            config.get_runbook_catalog(),
+            system_prompt_additions,
         )
         return
 
     messages = build_initial_ask_messages(
         console,
-        system_prompt_rendered,
         prompt,  # type: ignore
         include_file,
+        ai.tool_executor,
+        config.get_runbook_catalog(),
+        system_prompt_additions,
     )
 
     with tracer.start_trace(
diff --git a/holmes/plugins/prompts/generic_ask.jinja2 b/holmes/plugins/prompts/generic_ask.jinja2
index 07a0fe7670..fb7d9b35c8 100644
--- a/holmes/plugins/prompts/generic_ask.jinja2
+++ b/holmes/plugins/prompts/generic_ask.jinja2
@@ -36,3 +36,7 @@ Relevant logs:
 ```
 
 Validation error led to unhandled Java exception causing a crash.
+
+{% if system_prompt_additions %}
+{{ system_prompt_additions }}
+{% endif %}
diff --git a/tests/core/test_prompt.py b/tests/core/test_prompt.py
new file mode 100644
index 0000000000..4e22096057
--- /dev/null
+++ b/tests/core/test_prompt.py
@@ -0,0 +1,185 @@
+import pytest
+from unittest.mock import Mock
+from rich.console import Console
+
+from holmes.core.prompt import (
+    build_initial_ask_messages,
+    append_file_to_user_prompt,
+    append_all_files_to_user_prompt,
+)
+
+
+@pytest.fixture
+def console():
+    return Console(force_terminal=False, force_jupyter=False)
+
+
+@pytest.fixture
+def mock_tool_executor():
+    tool_executor = Mock()
+    tool_executor.toolsets = []
+    return tool_executor
+
+
+def test_build_initial_ask_messages_basic(console, mock_tool_executor):
+    """Test basic message building without any optional parameters."""
+    messages = build_initial_ask_messages(
+        console,
+        "Test prompt",
+        None,
+        mock_tool_executor,
+        None,
+        None,
+    )
+
+    assert len(messages) == 2
+    assert messages[0]["role"] == "system"
+    assert messages[1]["role"] == "user"
+    assert messages[1]["content"] == "Test prompt"
+
+
+def test_build_initial_ask_messages_with_system_prompt_additions(
+    console, mock_tool_executor
+):
+    """Test message building with system prompt additions."""
+    system_additions = "Additional system instructions here."
+    messages = build_initial_ask_messages(
+        console,
+        "Test prompt",
+        None,
+        mock_tool_executor,
+        None,
+        system_additions,
+    )
+
+    assert len(messages) == 2
+    assert messages[0]["role"] == "system"
+    # Check for unique word from the system additions
+    assert "Additional" in messages[0]["content"]
+    assert messages[1]["role"] == "user"
+    assert messages[1]["content"] == "Test prompt"
+
+
+def test_build_initial_ask_messages_with_file(console, mock_tool_executor, tmp_path):
+    """Test message building with file attachment."""
+    # Create a temporary file
+    test_file = tmp_path / "test.txt"
+    test_file.write_text("File content here")
+
+    messages = build_initial_ask_messages(
+        console,
+        "Test prompt",
+        [test_file],
+        mock_tool_executor,
+        None,
+        None,
+    )
+
+    assert len(messages) == 2
+    assert messages[0]["role"] == "system"
+    assert messages[1]["role"] == "user"
+    assert "Test prompt" in messages[1]["content"]
+    assert "File content here" in messages[1]["content"]
+    # Check for file attachment markers
+    assert "<attached-file" in messages[1]["content"]
+    assert "test.txt" in messages[1]["content"]
+    assert "</attached-file>" in messages[1]["content"]
+
+
+def test_build_initial_ask_messages_with_runbooks(console, mock_tool_executor):
+    """Test message building with runbooks."""
+    runbooks = {"test_runbook": {"description": "Test runbook"}}
+
+    messages = build_initial_ask_messages(
+        console,
+        "Test prompt",
+        None,
+        mock_tool_executor,
+        runbooks,
+        None,
+    )
+
+    assert len(messages) == 2
+    assert messages[0]["role"] == "system"
+    # The runbook should be passed to the template context
+    assert messages[1]["role"] == "user"
+    assert messages[1]["content"] == "Test prompt"
+
+
+def test_build_initial_ask_messages_all_parameters(
+    console, mock_tool_executor, tmp_path
+):
+    """Test message building with all parameters."""
+    # Create a temporary file
+    test_file = tmp_path / "test.txt"
+    test_file.write_text("File content")
+
+    runbooks = {"test_runbook": {"description": "Test runbook"}}
+    system_additions = "Extra system instructions"
+
+    messages = build_initial_ask_messages(
+        console,
+        "Test prompt",
+        [test_file],
+        mock_tool_executor,
+        runbooks,
+        system_additions,
+    )
+
+    assert len(messages) == 2
+    assert messages[0]["role"] == "system"
+    # Check for unique word from system additions
+    assert "Extra" in messages[0]["content"]
+    assert messages[1]["role"] == "user"
+    assert "Test prompt" in messages[1]["content"]
+    assert "File content" in messages[1]["content"]
+
+
+def test_append_file_to_user_prompt(tmp_path):
+    """Test appending a single file to user prompt."""
+    test_file = tmp_path / "test.txt"
+    test_file.write_text("Test file content")
+
+    prompt = "Original prompt"
+    result = append_file_to_user_prompt(prompt, test_file)
+
+    assert "Original prompt" in result
+    assert "Test file content" in result
+    # Check for file attachment markers
+    assert "<attached-file" in result
+    assert "test.txt" in result
+    assert "</attached-file>" in result
+
+
+def test_append_all_files_to_user_prompt(console, tmp_path):
+    """Test appending multiple files to user prompt."""
+    # Create multiple test files
+    file1 = tmp_path / "file1.txt"
+    file1.write_text("Content 1")
+
+    file2 = tmp_path / "file2.txt"
+    file2.write_text("Content 2")
+
+    prompt = "Original prompt"
+    result = append_all_files_to_user_prompt(console, prompt, [file1, file2])
+
+    assert "Original prompt" in result
+    assert "Content 1" in result
+    assert "Content 2" in result
+    # Check for file attachment markers
+    assert "<attached-file" in result
+    assert "file1.txt" in result
+    assert "file2.txt" in result
+    assert result.count("</attached-file>") == 2
+
+
+def test_append_all_files_to_user_prompt_no_files(console):
+    """Test appending files when no files are provided."""
+    prompt = "Original prompt"
+    result = append_all_files_to_user_prompt(console, prompt, None)
+
+    assert result == "Original prompt"
+
+    # Also test with empty list
+    result = append_all_files_to_user_prompt(console, prompt, [])
+    assert result == "Original prompt"
diff --git a/tests/llm/test_ask_holmes.py b/tests/llm/test_ask_holmes.py
index ba8ebc652a..b13bdc41a1 100644
--- a/tests/llm/test_ask_holmes.py
+++ b/tests/llm/test_ask_holmes.py
@@ -25,8 +25,10 @@
     Evaluation,
     MockHelper,
     check_and_skip_test,
-    build_initial_ask_messages2,
 )
+
+from holmes.core.prompt import build_initial_ask_messages
+
 from tests.llm.utils.property_manager import (
     set_initial_properties,
     update_test_results,
@@ -163,6 +165,7 @@ def test_ask_holmes(
                     expected=test_case.expected_output,
                     dataset_record_id=test_case.id,
                     scores={},
+                    # metadata={"tags": test_case.tags},
                 )
         except Exception:
             pass  # Don't fail the test due to logging issues
@@ -227,6 +230,7 @@ def test_ask_holmes(
             dataset_record_id=test_case.id,
             scores=scores,
             metadata={"system_prompt": prompt},
+            # metadata={"tags": test_case.tags},
         )
 
     # Print tool calls summary
@@ -311,7 +315,7 @@ def ask_holmes(
 
                 runbook_catalog = load_runbook_catalog()
                 runbooks = runbook_catalog.model_dump() if runbook_catalog else {}
-            messages = build_initial_ask_messages2(
+            messages = build_initial_ask_messages(
                 console,
                 test_case.user_prompt,
                 None,
diff --git a/tests/llm/utils/test_case_utils.py b/tests/llm/utils/test_case_utils.py
index 4854ad2eb5..5b714b9ae9 100644
--- a/tests/llm/utils/test_case_utils.py
+++ b/tests/llm/utils/test_case_utils.py
@@ -10,17 +10,11 @@
 import pytest
 from pydantic import BaseModel, TypeAdapter
 from holmes.core.models import InvestigateRequest, WorkloadHealthRequest
-from holmes.core.prompt import (
-    append_all_files_to_user_prompt,
-    append_file_to_user_prompt,
-)
+from holmes.core.prompt import append_file_to_user_prompt
+
 from holmes.core.tool_calling_llm import ResourceInstructions
-from holmes.plugins.runbooks import RunbookCatalog
-from holmes.plugins.prompts import load_and_render_prompt
 from tests.llm.utils.constants import ALLOWED_EVAL_TAGS
 
-from rich.console import Console
-
 
 def read_file(file_path: Path):
     with open(file_path, "r", encoding="utf-8") as file:
@@ -285,43 +279,3 @@ def load_include_files(
             extra_prompt = append_file_to_user_prompt(extra_prompt, file_path)
 
     return extra_prompt
-
-
-# temporary until we merge https://github.com/robusta-dev/holmesgpt/pull/727
-def build_initial_ask_messages2(
-    console: Console,
-    initial_user_prompt: str,
-    file_paths: Optional[List[Path]],
-    tool_executor: Any,  # ToolExecutor type
-    runbooks: Union[RunbookCatalog, Dict, None] = None,
-) -> List[Dict]:
-    """Build the initial messages for the AI call.
-
-    Args:
-        console: Rich console for output
-        initial_user_prompt: The user's prompt
-        file_paths: Optional list of files to include
-        tool_executor: The tool executor with available toolsets
-        runbooks: Optional runbook catalog
-    """
-    # Load and render system prompt internally
-    system_prompt_template = "builtin://generic_ask.jinja2"
-    template_context = {
-        "toolsets": tool_executor.toolsets,
-        "runbooks": runbooks or {},
-    }
-    system_prompt_rendered = load_and_render_prompt(
-        system_prompt_template, template_context
-    )
-
-    # Append files to user prompt
-    user_prompt_with_files = append_all_files_to_user_prompt(
-        console, initial_user_prompt, file_paths
-    )
-
-    messages = [
-        {"role": "system", "content": system_prompt_rendered},
-        {"role": "user", "content": user_prompt_with_files},
-    ]
-
-    return messages