diff --git a/env_example b/env_example
index 3207596..6ff76ee 100644
--- a/env_example
+++ b/env_example
@@ -1,5 +1,6 @@
 GITHUB_TOKEN=token123
 GITHUB_REPOSITORY=owner/repo
 RAW_PROMPT_URL=https://raw.githubusercontent.com/johndoe6345789/metabuilder/main/getonwithit.prompt.yml
+PROMPT_PATH=prompt.yml
 GITHUB_MODELS_ENDPOINT=https://models.github.ai/inference
 APP_LANG=en # Supported: en, es, fr, nl, pirate
diff --git a/prompt.yml b/prompt.yml
new file mode 100644
index 0000000..cb92ac5
--- /dev/null
+++ b/prompt.yml
@@ -0,0 +1,16 @@
+messages:
+  - role: system
+    content: >-
+      Use software architecture best practices, unit testing, e2e testing,
+      linting.
+  - role: user
+    content: >-
+      Implement features found in ROADMAP.md and README.md, keep both files up
+      to date. Write playwright and unit tests. Follow code style. Use JSON
+      schema but its mostly correct so leave the schema alone. Index repository
+      and make notes. Align styling with old/ folder while sticking with plain
+      SASS files. Work towards MVP. Run linter and fix lint issues. Ensure
+      development DBAL is aligned with C++ production DBAL. Tables, Prisma
+      schemas and types should be generated using DBAL. C++ CLI frontend can
+      interface with DBAL and make life a bit easier. Implement any stub code.
+model: openai/gpt-4o
diff --git a/src/autometabuilder/main.py b/src/autometabuilder/main.py
index b58e3a4..dd36ee7 100644
--- a/src/autometabuilder/main.py
+++ b/src/autometabuilder/main.py
@@ -12,6 +12,7 @@ from .github_integration import GitHubIntegration, get_repo_name_from_env
 
 load_dotenv()
 
+DEFAULT_PROMPT_PATH = "prompt.yml"
 DEFAULT_RAW_PROMPT_URL = (
     "https://raw.githubusercontent.com/johndoe6345789/"
     "metabuilder/main/getonwithit.prompt.yml"
@@ -19,8 +20,16 @@ DEFAULT_RAW_PROMPT_URL = (
 DEFAULT_ENDPOINT = "https://models.github.ai/inference"
 
 
-def load_prompt_yaml(url: str, token: str) -> dict:
-    """Load prompt configuration from a remote YAML file."""
+def load_prompt_yaml(token: str) -> dict:
+    """Load prompt configuration from local file or remote URL."""
+    # Try local file first
+    local_path = os.environ.get("PROMPT_PATH", DEFAULT_PROMPT_PATH)
+    if os.path.exists(local_path):
+        with open(local_path, "r", encoding="utf-8") as f:
+            return yaml.safe_load(f)
+
+    # Fallback to remote URL
+    url = os.environ.get("RAW_PROMPT_URL", DEFAULT_RAW_PROMPT_URL)
     headers = {"Authorization": f"Bearer {token}"}
     r = requests.get(url, headers=headers, timeout=30)
     r.raise_for_status()
@@ -94,9 +103,7 @@ def main():
         api_key=token,
     )
 
-    prompt = load_prompt_yaml(
-        os.environ.get("RAW_PROMPT_URL", DEFAULT_RAW_PROMPT_URL), token
-    )
+    prompt = load_prompt_yaml(token)
 
     # Load tools for SDLC operations from JSON file
     tools_path = os.path.join(os.path.dirname(__file__), "tools.json")