v3.3.1: Strict Coder - no sandbox exploration commands

Dockerfile

@@ -1,5 +1,5 @@
 # HuggingFace Spaces Dockerfile for CodePilot
-# BUILD_VERSION: 9 (v3.3.0 clarifying questions)
+# BUILD_VERSION: 10 (v3.3.1 coder strict - no sandbox exploration)
 FROM python:3.11-slim
 
 # Set working directory

chainlit_app.py

@@ -20,8 +20,8 @@ from concurrent.futures import ThreadPoolExecutor
 # ============================================================
 # STARTUP VERSION CHECK - Change this to detect if rebuild worked
 # ============================================================
-APP_VERSION = "3.3.0-clarify"
-BUILD_ID = "2024-12-19-v8"
+APP_VERSION = "3.3.1-coder-strict"
+BUILD_ID = "2024-12-19-v9"
 print("=" * 60)
 print(f"[STARTUP] CodePilot Chainlit App")
 print(f"[STARTUP] APP_VERSION: {APP_VERSION}")

codepilot/agents/coder_agent.py

@@ -21,51 +21,34 @@ from typing import Dict, Any, Optional
 import json
 
 
-# Coder's specialized system prompt (v3.2 - no search, no list_files, uses exploration context)
-CODER_SYSTEM_PROMPT = """You are an expert software engineer and implementation specialist.
+# Coder's specialized system prompt (v3.3 - no exploration, write code immediately)
+CODER_SYSTEM_PROMPT = """You are an expert software engineer. Your ONLY job is to WRITE CODE.
+
+=== CRITICAL RULES ===
+1. START WRITING CODE IMMEDIATELY with write_file
+2. DO NOT run ls, find, pwd, or any exploration commands
+3. DO NOT try to access /tmp/codepilot_repos/ from sandbox - IT WILL FAIL
+4. Use file paths EXACTLY as shown in the exploration results
+5. The sandbox is ISOLATED - it cannot see the cloned repo
+
+=== YOUR WORKFLOW ===
+1. Look at the EXPLORATION RESULTS - they have all the file paths you need
+2. Look at the PLAN - it tells you exactly what to create
+3. Use write_file to create/modify files in the repo
+4. (Optional) Test your code by uploading to sandbox with relative paths
 
-Your ONLY job is to write code that implements the given plan. You do NOT explore or search.
-
-=== CRITICAL: USE THE PROVIDED CONTEXT ===
-The Explorer agent has ALREADY searched the codebase for you. All file paths and code patterns are in the EXPLORATION RESULTS below.
-
-DO NOT:
-- Navigate directories (no list_files)
-- Search for files (no searching)
-- Explore the codebase
-
-DO:
-- Use the exact file paths from exploration results
-- Start writing code immediately
-- Follow the plan step by step
+=== TOOLS ===
+- write_file: CREATE OR MODIFY FILES - use this immediately!
+- get_code_chunk: Read a specific function if you need to see existing code
+- upload_to_sandbox: Upload YOUR code for testing (use relative paths like "app.py")
+- run_command_in_sandbox: Run YOUR uploaded code (like "python app.py")
 
-=== WORKFLOW ===
-1. Read the exploration results - they contain all file paths you need
-2. If modifying existing code: use get_code_chunk to read the specific function
-3. Write your changes with write_file using paths from exploration
-4. Test in sandbox if needed
+=== EXAMPLE ===
+If plan says "create examples/health_check/app.py":
+✅ CORRECT: write_file(path="/tmp/codepilot_repos/flask_xxx/examples/health_check/app.py", content="...")
+❌ WRONG: run_command_in_sandbox("ls /tmp/codepilot_repos/...")  <-- sandbox can't access this!
 
-=== TOOLS ===
-- get_file_outline: See file structure (use if unsure about a file)
-- get_code_chunk: Read ONE specific function/class
-- read_file: Read entire file (only when rewriting whole file)
-- write_file: Create or modify files
-- upload_to_sandbox: Upload files for testing
-- run_command_in_sandbox: Run tests in sandbox
-- execute_in_sandbox: Execute Python snippets
-
-=== SANDBOX WORKFLOW ===
-When testing in sandbox:
-1. Upload with RELATIVE path: upload_to_sandbox(path="file.py", content=code)
-2. Run with RELATIVE path: run_command_in_sandbox(command="python file.py")
-3. The sandbox CANNOT access /tmp/codepilot_repos/ - use simple filenames!
-
-Your code should be:
-- Clean (follow existing code style)
-- Minimal (only change what's necessary)
-- Follow the plan exactly
-
-START CODING IMMEDIATELY - do not explore!
+DO NOT EXPLORE. WRITE CODE NOW.
 """
 
 
@@ -94,17 +77,16 @@ class CoderAgent:
 
         self.conversation = ConversationManager()
 
-        # v3.2: Removed list_files - Explorer provides all paths needed
-        # Coder only needs: read, write, and sandbox tools
+        # v3.3: Minimal tools - focus on writing code
+        # Removed sandbox commands that were being misused for exploration
         self.allowed_tools = [
-            "get_file_outline",     # Get file structure without full code
-            "get_code_chunk",       # Extract specific function/class by name
-            "read_file",            # Full file contents (use sparingly)
-            "write_file",           # Create or modify files
-            # "list_files" REMOVED - use exploration context instead
-            "upload_to_sandbox",    # Upload files for testing
-            "run_command_in_sandbox",  # Run tests in sandbox
-            "execute_in_sandbox"    # Execute Python snippets
+            "write_file",           # PRIMARY TOOL - create or modify files
+            "get_code_chunk",       # Read specific function/class if needed
+            "read_file",            # Read entire file (only when rewriting)
+            "upload_to_sandbox",    # Upload code for testing
+            "execute_in_sandbox"    # Execute Python snippets for testing
+            # REMOVED: run_command_in_sandbox - was being misused for ls/find
+            # REMOVED: get_file_outline - Coder should use exploration context
         ]
 
     def run(

codepilot/agents/orchestrator.py

@@ -9,7 +9,7 @@ The orchestrator is the "brain" that:
 """
 
 # VERSION CHECK - If you see this, new code is running!
-ORCHESTRATOR_VERSION = "3.3.0-clarify"
+ORCHESTRATOR_VERSION = "3.3.1-coder-strict"
 print(f"[ORCHESTRATOR] ========== LOADING VERSION {ORCHESTRATOR_VERSION} ==========")
 
 from enum import Enum