From 07cd9a1f321dd6aadb84f20ee3cc337491f9a483 Mon Sep 17 00:00:00 2001
From: Test User <test@example.com>
Date: Thu, 15 Jan 2026 08:16:15 +0100
Subject: [PATCH] story4.3

---
 apps/backend/core/executors/full_auto.py | 431 ++++++++++++++++++-
 tests/executors/test_full_auto.py        | 504 +++++++++++++++++++++++
 2 files changed, 934 insertions(+), 1 deletion(-)

diff --git a/apps/backend/core/executors/full_auto.py b/apps/backend/core/executors/full_auto.py
index 894b972f..c59a747b 100644
--- a/apps/backend/core/executors/full_auto.py
+++ b/apps/backend/core/executors/full_auto.py
@@ -2,6 +2,7 @@
 
 Story Reference: Story 4.1 - Implement Full Auto Task Executor
 Story Reference: Story 4.2 - Implement Planning Phase Execution
+Story Reference: Story 4.3 - Implement Coding Phase Execution
 
 This module provides the FullAutoExecutor class that executes all methodology
 phases without user intervention. In Full Auto mode, the system executes
@@ -11,6 +12,7 @@ continuously.
 Architecture Source: architecture.md#Task-Execution
 """
 
+import asyncio
 import json
 import logging
 import time
@@ -18,7 +20,7 @@ from dataclasses import dataclass, field
 from datetime import datetime
 from enum import StrEnum
 from pathlib import Path
-from typing import TYPE_CHECKING, Literal
+from typing import TYPE_CHECKING, Any, Literal
 
 # fcntl is POSIX only - use for file locking on Unix/macOS
 try:
@@ -180,6 +182,17 @@ class FullAutoExecutor:
         self._phases: list[Phase] = []
         self._current_state: TaskState | None = None
 
+        # Story 4.3: Initialize spec_dir and worktree_path for coding phase
+        # These can be set from task_config.metadata or explicitly via properties
+        spec_dir_str = (
+            task_config.metadata.get("spec_dir") if task_config.metadata else None
+        )
+        self._spec_dir: Path | None = Path(spec_dir_str) if spec_dir_str else None
+        worktree_str = (
+            task_config.metadata.get("worktree_path") if task_config.metadata else None
+        )
+        self._worktree_path: Path | None = Path(worktree_str) if worktree_str else None
+
         # Task-specific logger for filtering logs by task ID.
         # Logger name format: "apps.backend.core.executors.full_auto.{task_id}"
         # This allows log filtering with: logging.getLogger("...full_auto.task-123")
@@ -719,3 +732,419 @@ class FullAutoExecutor:
             "qa_validation": TaskState.VALIDATION_COMPLETE,
         }
         return complete_state_map.get(phase_id)
+
+    # =========================================================================
+    # Story 4.3: Coding Phase Execution Methods
+    # =========================================================================
+
+    def load_implementation_plan(self, spec_dir: Path) -> list[dict]:
+        """Load subtasks from implementation_plan.json.
+
+        Story Reference: Story 4.3 Task 1 - Load implementation plan
+
+        Args:
+            spec_dir: Directory containing implementation_plan.json
+
+        Returns:
+            List of subtask dictionaries
+
+        Raises:
+            FileNotFoundError: If implementation_plan.json doesn't exist
+            json.JSONDecodeError: If the file contains invalid JSON
+        """
+        plan_path = spec_dir / "implementation_plan.json"
+        if not plan_path.exists():
+            raise FileNotFoundError(f"Implementation plan not found: {plan_path}")
+
+        plan_data = json.loads(plan_path.read_text())
+        return plan_data.get("subtasks", [])
+
+    def get_pending_subtasks(self, spec_dir: Path) -> list[dict]:
+        """Get subtasks that haven't been completed yet.
+
+        Story Reference: Story 4.3 Task 1 - Determine execution order
+
+        Args:
+            spec_dir: Directory containing implementation_plan.json
+
+        Returns:
+            List of subtasks with status "pending"
+        """
+        all_subtasks = self.load_implementation_plan(spec_dir)
+        return [s for s in all_subtasks if s.get("status") == "pending"]
+
+    async def execute_subtask(self, subtask: dict) -> dict:
+        """Execute a single subtask.
+
+        Story Reference: Story 4.3 Task 2 - Implement subtask loop
+
+        This method:
+        1. Marks subtask as in_progress
+        2. Delegates to methodology runner
+        3. Updates subtask status on completion/failure
+
+        Args:
+            subtask: The subtask dictionary to execute
+
+        Returns:
+            Result dictionary with "success" key and optional error
+        """
+        subtask_id = subtask.get("id", "unknown")
+        self._log_info(f"Executing subtask: {subtask_id}", phase_id="coding")
+
+        # Mark as in_progress in the plan file
+        self._update_subtask_status(subtask_id, "in_progress")
+
+        try:
+            # Execute via methodology runner
+            result = await self._execute_subtask_impl(subtask)
+
+            if result.get("success"):
+                self._update_subtask_status(subtask_id, "completed")
+            else:
+                self._update_subtask_status(subtask_id, "failed")
+
+            return result
+
+        except Exception as e:
+            self._update_subtask_status(subtask_id, "failed")
+            return {"success": False, "error": str(e)}
+
+    async def _execute_subtask_impl(
+        self,
+        subtask: dict,
+        attempt: int = 1,
+        recovery_context: str | None = None,
+    ) -> dict:
+        """Internal implementation of subtask execution.
+
+        Story Reference: Story 4.3 Task 2 - Execute via methodology runner
+
+        This method is the core execution logic that can be mocked for testing.
+
+        Args:
+            subtask: The subtask to execute
+            attempt: Current attempt number (1-based)
+            recovery_context: Optional context from previous failed attempts
+
+        Returns:
+            Result dictionary with "success" and optional "error" keys
+        """
+        # Default implementation - delegates to runner.execute_phase
+        # In real usage, this would invoke the coder agent
+        try:
+            if hasattr(self.runner, "execute_subtask"):
+                result = self.runner.execute_subtask(subtask, attempt, recovery_context)
+                return {"success": result.success, "error": result.error}
+            return {"success": True}
+        except Exception as e:
+            return {"success": False, "error": str(e)}
+
+    def _update_subtask_status(self, subtask_id: str, status: str) -> None:
+        """Update a subtask's status in the implementation plan.
+
+        Story Reference: Story 4.3 Task 2 - Update status
+
+        Args:
+            subtask_id: ID of the subtask to update
+            status: New status value ("in_progress", "completed", "failed")
+        """
+        if not self._spec_dir:
+            return
+
+        plan_path = self._spec_dir / "implementation_plan.json"
+        if not plan_path.exists():
+            return
+
+        try:
+            plan_data = json.loads(plan_path.read_text())
+            subtasks = plan_data.get("subtasks", [])
+
+            for subtask in subtasks:
+                if subtask.get("id") == subtask_id:
+                    subtask["status"] = status
+                    break
+
+            plan_path.write_text(json.dumps(plan_data, indent=2))
+        except Exception as e:
+            self._log_warning(f"Failed to update subtask status: {e}")
+
+    async def execute_coding_phase(self) -> dict[str, Any]:
+        """Execute the full coding phase by running all subtasks.
+
+        Story Reference: Story 4.3 Task 2 - Iterate through subtasks
+
+        Returns:
+            Result dictionary with keys:
+            - status: "coding_complete", "failed", "escalated", or "completed"
+            - subtasks_executed: Number of successfully executed subtasks
+            - subtask: ID of failed subtask (if failed/escalated)
+            - error: Error message (if failed/escalated)
+        """
+        if not self._spec_dir:
+            self._log_warning("No spec_dir set for coding phase")
+            return {"status": "failed", "error": "No spec_dir configured"}
+
+        try:
+            pending_subtasks = self.get_pending_subtasks(self._spec_dir)
+        except FileNotFoundError:
+            self._log_warning("No implementation plan found")
+            return {"status": "completed", "subtasks_executed": 0}
+
+        total_subtasks = len(pending_subtasks)
+        if total_subtasks == 0:
+            self._log_info("No pending subtasks - coding phase complete")
+            self.update_task_state(TaskState.CODING_COMPLETE)
+            return {"status": "coding_complete", "subtasks_executed": 0}
+
+        self._log_info(f"Starting coding phase with {total_subtasks} subtasks")
+        self.update_task_state(TaskState.CODING)
+
+        executed = 0
+        for index, subtask in enumerate(pending_subtasks):
+            subtask_id = subtask.get("id", "unknown")
+
+            # Emit progress on start
+            self.emit_subtask_progress(index, total_subtasks, subtask, "started")
+
+            # Execute with retry logic
+            result = await self.execute_subtask_with_retry(subtask)
+
+            if result.get("success"):
+                executed += 1
+                # Update subtask status to completed in plan file
+                self._update_subtask_status(subtask_id, "completed")
+                # Emit progress on complete
+                self.emit_subtask_progress(index, total_subtasks, subtask, "completed")
+                # Commit changes after successful subtask
+                self.commit_subtask_changes(self.format_commit_message(subtask))
+            elif result.get("escalated"):
+                # Escalation - preserve partial progress before stopping
+                self._update_subtask_status(subtask_id, "failed")
+                self.emit_subtask_progress(index, total_subtasks, subtask, "failed")
+                self.update_task_state(TaskState.ESCALATED)
+                self._log_error(
+                    f"Subtask {subtask_id} escalated after max retries",
+                    phase_id="coding",
+                )
+                return {
+                    "status": "escalated",
+                    "subtask": subtask_id,
+                    "subtasks_executed": executed,
+                    "error": result.get("error"),
+                }
+            else:
+                # Non-escalated failure (unexpected) - stop execution and mark failed
+                self._update_subtask_status(subtask_id, "failed")
+                self.emit_subtask_progress(index, total_subtasks, subtask, "failed")
+                self.update_task_state(TaskState.FAILED)
+                error_msg = result.get("error", "Unknown failure")
+                self._log_error(
+                    f"Subtask {subtask_id} failed: {error_msg}",
+                    phase_id="coding",
+                )
+                return {
+                    "status": "failed",
+                    "subtask": subtask_id,
+                    "subtasks_executed": executed,
+                    "error": error_msg,
+                }
+
+        self._log_info(f"Coding phase complete: {executed}/{total_subtasks} subtasks")
+        self.update_task_state(TaskState.CODING_COMPLETE)
+        return {"status": "coding_complete", "subtasks_executed": executed}
+
+    def emit_subtask_progress(
+        self,
+        index: int,
+        total: int,
+        subtask: dict,
+        status: str,
+    ) -> None:
+        """Emit progress event for a subtask.
+
+        Story Reference: Story 4.3 Task 3 - Per-subtask progress reporting
+
+        Args:
+            index: Current subtask index (0-based)
+            total: Total number of subtasks
+            subtask: The subtask dictionary
+            status: Status string ("started", "completed", "failed")
+        """
+        percentage = self.calculate_subtask_percentage(index, total)
+        subtask_id = subtask.get("id", "unknown")
+        subtask_title = subtask.get("title", subtask.get("description", "subtask"))
+
+        message = f"Subtask {status}: {subtask_title}"
+
+        progress_status = (
+            ProgressStatus.STARTED
+            if status == "started"
+            else (
+                ProgressStatus.COMPLETED
+                if status == "completed"
+                else ProgressStatus.FAILED
+            )
+        )
+
+        self._emit_progress_event(
+            phase_id="coding",
+            status=progress_status,
+            message=message,
+            percentage=percentage,
+        )
+
+    def calculate_subtask_percentage(self, completed: int, total: int) -> float:
+        """Calculate progress percentage for subtask execution.
+
+        Story Reference: Story 4.3 Task 3 - Calculate percentage
+
+        The coding phase spans 35-75% of overall progress (40% range).
+        This method calculates where within that range we are.
+
+        Args:
+            completed: Number of completed subtasks
+            total: Total number of subtasks
+
+        Returns:
+            Percentage as float (35.0 to 75.0)
+        """
+        if total == 0:
+            return 35.0  # Start of coding phase
+
+        # Coding phase is 35-75% (40% range)
+        base_percentage = 35.0
+        range_percentage = 40.0
+        return base_percentage + (completed / total) * range_percentage
+
+    def commit_subtask_changes(self, message: str) -> bool:
+        """Commit changes after subtask completion.
+
+        Story Reference: Story 4.3 Task 4 - Git commit integration
+
+        Args:
+            message: Commit message
+
+        Returns:
+            True if commit succeeded, False otherwise
+        """
+        import subprocess
+
+        if not self._worktree_path:
+            self._log_warning("No worktree path configured for commits")
+            return False
+
+        try:
+            # Stage all changes including deletions (git add -A as per Dev Notes)
+            result = subprocess.run(
+                ["git", "add", "-A"],
+                cwd=self._worktree_path,
+                capture_output=True,
+                text=True,
+            )
+
+            if result.returncode != 0:
+                self._log_warning(f"Git add failed: {result.stderr}")
+                return False
+
+            # Commit
+            result = subprocess.run(
+                ["git", "commit", "-m", message],
+                cwd=self._worktree_path,
+                capture_output=True,
+                text=True,
+            )
+
+            if result.returncode != 0:
+                if (
+                    "nothing to commit" in result.stdout
+                    or "nothing to commit" in result.stderr
+                ):
+                    return True  # No changes to commit
+                self._log_warning(f"Git commit failed: {result.stderr}")
+                return False
+
+            self._log_info(f"Committed changes: {message}")
+            return True
+
+        except Exception as e:
+            self._log_warning(f"Git operation failed: {e}")
+            return False
+
+    def format_commit_message(self, subtask: dict) -> str:
+        """Format a commit message for a subtask.
+
+        Story Reference: Story 4.3 Task 4 - Meaningful commit messages
+
+        Args:
+            subtask: The completed subtask
+
+        Returns:
+            Formatted commit message string
+        """
+        subtask_id = subtask.get("id", "unknown")
+        title = subtask.get("title", subtask.get("description", "subtask"))
+        return f"Implement: {title} (#{subtask_id})"
+
+    async def execute_subtask_with_retry(
+        self,
+        subtask: dict,
+        max_retries: int = 3,
+    ) -> dict:
+        """Execute a subtask with retry logic.
+
+        Story Reference: Story 4.3 Task 5 - Retry logic
+
+        Implements exponential backoff with recovery context passed
+        on retry attempts.
+
+        Args:
+            subtask: The subtask to execute
+            max_retries: Maximum number of attempts (default 3)
+
+        Returns:
+            Result dictionary with "success", "error", and potentially "escalated"
+        """
+
+        last_error = None
+        recovery_context = None
+
+        for attempt in range(1, max_retries + 1):
+            self._log_info(
+                f"Executing subtask {subtask.get('id')} (attempt {attempt}/{max_retries})",
+                phase_id="coding",
+            )
+
+            result = await self._execute_subtask_impl(
+                subtask,
+                attempt=attempt,
+                recovery_context=recovery_context,
+            )
+
+            if result.get("success"):
+                return result
+
+            last_error = result.get("error", "Unknown error")
+            recovery_context = f"Previous attempt {attempt} failed: {last_error}"
+
+            # Exponential backoff (1s, 2s, 4s)
+            if attempt < max_retries:
+                backoff = 2 ** (attempt - 1)
+                self._log_info(
+                    f"Retry in {backoff}s due to: {last_error}",
+                    phase_id="coding",
+                )
+                await asyncio.sleep(backoff)
+
+        # All retries exhausted - escalate
+        self._log_error(
+            f"Subtask {subtask.get('id')} failed after {max_retries} attempts: {last_error}",
+            phase_id="coding",
+        )
+        self.update_task_state(TaskState.ESCALATED)
+
+        return {
+            "success": False,
+            "error": f"Escalated after {max_retries} failed attempts: {last_error}",
+            "escalated": True,
+        }
diff --git a/tests/executors/test_full_auto.py b/tests/executors/test_full_auto.py
index 3a348483..098d14ac 100644
--- a/tests/executors/test_full_auto.py
+++ b/tests/executors/test_full_auto.py
@@ -1019,3 +1019,507 @@ class TestArtifactVerificationIntegration:
 
             # Phase executed but verification should be part of the flow
             assert "planning" in runner._executed_phases
+
+
+# =============================================================================
+# Story 4.3: Coding Phase Execution Tests
+# =============================================================================
+
+
+class TestCodingPhaseSubtaskLoading:
+    """Test loading implementation plan for coding phase (Story 4.3 Task 1)."""
+
+    def test_load_implementation_plan(self, executor):
+        """Test loading subtasks from implementation_plan.json."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            spec_dir = Path(tmpdir)
+            plan_data = {
+                "spec_id": "test-spec",
+                "subtasks": [
+                    {"id": "1", "title": "First subtask", "status": "pending"},
+                    {"id": "2", "title": "Second subtask", "status": "pending"},
+                    {"id": "3", "title": "Third subtask", "status": "completed"},
+                ],
+            }
+            import json
+
+            (spec_dir / "implementation_plan.json").write_text(json.dumps(plan_data))
+
+            subtasks = executor.load_implementation_plan(spec_dir)
+
+            assert len(subtasks) == 3
+            assert subtasks[0]["id"] == "1"
+            assert subtasks[1]["title"] == "Second subtask"
+
+    def test_load_implementation_plan_missing_file(self, executor):
+        """Test that missing plan file raises appropriate error."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            spec_dir = Path(tmpdir)
+
+            with pytest.raises(FileNotFoundError):
+                executor.load_implementation_plan(spec_dir)
+
+    def test_load_implementation_plan_invalid_json(self, executor):
+        """Test that invalid JSON raises appropriate error."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            spec_dir = Path(tmpdir)
+            (spec_dir / "implementation_plan.json").write_text("invalid json")
+
+            with pytest.raises(Exception):  # Could be json.JSONDecodeError
+                executor.load_implementation_plan(spec_dir)
+
+    def test_get_pending_subtasks(self, executor):
+        """Test filtering to get only pending subtasks."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            spec_dir = Path(tmpdir)
+            plan_data = {
+                "subtasks": [
+                    {"id": "1", "title": "First", "status": "completed"},
+                    {"id": "2", "title": "Second", "status": "pending"},
+                    {"id": "3", "title": "Third", "status": "in_progress"},
+                    {"id": "4", "title": "Fourth", "status": "pending"},
+                ],
+            }
+            import json
+
+            (spec_dir / "implementation_plan.json").write_text(json.dumps(plan_data))
+
+            pending = executor.get_pending_subtasks(spec_dir)
+
+            assert len(pending) == 2
+            assert pending[0]["id"] == "2"
+            assert pending[1]["id"] == "4"
+
+
+class TestCodingPhaseSubtaskExecution:
+    """Test subtask execution loop (Story 4.3 Task 2)."""
+
+    @pytest.mark.asyncio
+    async def test_execute_subtask_updates_status(self, executor):
+        """Test that executing a subtask updates its status in the plan."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            spec_dir = Path(tmpdir)
+            plan_data = {
+                "subtasks": [
+                    {"id": "1", "title": "Test subtask", "status": "pending"},
+                ],
+            }
+            import json
+
+            plan_file = spec_dir / "implementation_plan.json"
+            plan_file.write_text(json.dumps(plan_data))
+            executor._spec_dir = spec_dir
+
+            # Execute the subtask
+            subtask = plan_data["subtasks"][0]
+            result = await executor.execute_subtask(subtask)
+
+            # Reload plan and check status updated
+            updated_plan = json.loads(plan_file.read_text())
+            # Status should be either "completed" (success) or "failed"
+            assert updated_plan["subtasks"][0]["status"] in ["completed", "in_progress", "failed"]
+
+    @pytest.mark.asyncio
+    async def test_execute_all_subtasks_in_sequence(self, executor):
+        """Test that all subtasks are executed in sequence."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            spec_dir = Path(tmpdir)
+            plan_data = {
+                "subtasks": [
+                    {"id": "1", "title": "First", "status": "pending"},
+                    {"id": "2", "title": "Second", "status": "pending"},
+                    {"id": "3", "title": "Third", "status": "pending"},
+                ],
+            }
+            import json
+
+            plan_file = spec_dir / "implementation_plan.json"
+            plan_file.write_text(json.dumps(plan_data))
+            executor._spec_dir = spec_dir
+            executor._task_dir = Path(tmpdir)
+
+            # Track execution order
+            executed_ids = []
+
+            async def mock_execute_impl(subtask, attempt=1, recovery_context=None):
+                executed_ids.append(subtask["id"])
+                return {"success": True}
+
+            # Patch the implementation method
+            with patch.object(executor, "_execute_subtask_impl", side_effect=mock_execute_impl):
+                await executor.execute_coding_phase()
+
+            # All subtasks should be executed in order
+            assert executed_ids == ["1", "2", "3"]
+
+
+class TestCodingPhaseProgressReporting:
+    """Test per-subtask progress reporting (Story 4.3 Task 3)."""
+
+    @pytest.mark.asyncio
+    async def test_emit_progress_on_subtask_start(self, executor, mock_progress_service):
+        """Test that progress is emitted when subtask starts."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            spec_dir = Path(tmpdir)
+            plan_data = {
+                "subtasks": [
+                    {"id": "1", "title": "Test subtask", "status": "pending"},
+                ],
+            }
+            import json
+
+            (spec_dir / "implementation_plan.json").write_text(json.dumps(plan_data))
+            executor._spec_dir = spec_dir
+            executor.context.progress = mock_progress_service
+
+            executor.emit_subtask_progress(0, 1, plan_data["subtasks"][0], "started")
+
+            # Should have emitted a started event
+            assert len(mock_progress_service.events) >= 1
+            last_event = mock_progress_service.events[-1]
+            assert "started" in last_event.message.lower() or last_event.status == "started"
+
+    @pytest.mark.asyncio
+    async def test_emit_progress_on_subtask_complete(self, executor, mock_progress_service):
+        """Test that progress is emitted when subtask completes."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            spec_dir = Path(tmpdir)
+            plan_data = {
+                "subtasks": [
+                    {"id": "1", "title": "Test subtask", "status": "pending"},
+                ],
+            }
+            import json
+
+            (spec_dir / "implementation_plan.json").write_text(json.dumps(plan_data))
+            executor._spec_dir = spec_dir
+            executor.context.progress = mock_progress_service
+
+            executor.emit_subtask_progress(0, 1, plan_data["subtasks"][0], "completed")
+
+            # Should have emitted a completed event
+            assert len(mock_progress_service.events) >= 1
+
+    @pytest.mark.asyncio
+    async def test_progress_percentage_calculation(self, executor, mock_progress_service):
+        """Test that percentage is calculated based on subtask count."""
+        # Coding phase is 35-75% of overall progress (40% range)
+        # 1 of 3 subtasks = 35 + (1/3)*40 = 35 + 13.33 = 48.33%
+
+        percentage = executor.calculate_subtask_percentage(1, 3)
+
+        # Should be around 48%
+        assert 45 <= percentage <= 55
+
+
+class TestCodingPhaseGitCommit:
+    """Test git commit integration (Story 4.3 Task 4)."""
+
+    @pytest.mark.asyncio
+    async def test_commit_after_subtask(self, executor):
+        """Test that changes are committed using git add -A to include deletions."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            executor._worktree_path = Path(tmpdir)
+
+            # Mock subprocess.run
+            with patch("subprocess.run") as mock_run:
+                mock_run.return_value.returncode = 0
+
+                executor.commit_subtask_changes("Implement: Test subtask")
+
+                # Verify git add -A is used (not git add .)
+                add_call = mock_run.call_args_list[0]
+                assert add_call[0][0] == ["git", "add", "-A"], (
+                    f"Expected 'git add -A', got {add_call[0][0]}"
+                )
+
+                # Verify commit call uses the message
+                commit_call = mock_run.call_args_list[1]
+                assert commit_call[0][0][:3] == ["git", "commit", "-m"]
+                assert "Implement: Test subtask" in commit_call[0][0][3]
+
+    @pytest.mark.asyncio
+    async def test_commit_uses_worktree_path(self, executor):
+        """Test that commits are made to the correct worktree directory."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            worktree_path = Path(tmpdir) / "worktree"
+            worktree_path.mkdir()
+            executor._worktree_path = worktree_path
+
+            with patch("subprocess.run") as mock_run:
+                mock_run.return_value.returncode = 0
+
+                executor.commit_subtask_changes("Test message")
+
+                # Verify cwd parameter is the worktree path
+                for call in mock_run.call_args_list:
+                    assert call[1].get("cwd") == worktree_path, (
+                        f"Expected cwd={worktree_path}, got {call[1].get('cwd')}"
+                    )
+
+    def test_commit_message_format(self, executor):
+        """Test that commit messages follow expected format."""
+        subtask = {"id": "1", "title": "Add user authentication"}
+
+        message = executor.format_commit_message(subtask)
+
+        assert "Implement:" in message or "Add user authentication" in message
+
+
+class TestCodingPhaseRetryLogic:
+    """Test retry logic for failed subtasks (Story 4.3 Task 5)."""
+
+    @pytest.mark.asyncio
+    async def test_retry_on_failure(self, executor):
+        """Test that failed subtask is retried up to 3 times."""
+        attempts = []
+
+        async def failing_execute(subtask, attempt=1, recovery_context=None):
+            attempts.append(attempt)
+            if len(attempts) < 3:
+                return {"success": False, "error": "Simulated failure"}
+            return {"success": True}
+
+        with patch.object(executor, "_execute_subtask_impl", side_effect=failing_execute):
+            subtask = {"id": "1", "title": "Test", "status": "pending"}
+            result = await executor.execute_subtask_with_retry(subtask)
+
+        assert len(attempts) == 3
+        assert result["success"] is True
+
+    @pytest.mark.asyncio
+    async def test_exponential_backoff(self, executor):
+        """Test that retries use exponential backoff with exact delay values."""
+        async def always_fail(subtask, attempt=1, recovery_context=None):
+            return {"success": False, "error": "Always fails"}
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            executor._task_dir = Path(tmpdir)
+            with patch.object(executor, "_execute_subtask_impl", side_effect=always_fail):
+                with patch("apps.backend.core.executors.full_auto.asyncio.sleep") as mock_sleep:
+                    subtask = {"id": "1", "title": "Test", "status": "pending"}
+                    await executor.execute_subtask_with_retry(subtask, max_retries=3)
+
+                    # Should have called sleep with exponential delays: 1, 2 (between 3 attempts)
+                    # Formula: 2 ** (attempt - 1) => 2^0=1, 2^1=2
+                    assert mock_sleep.call_count == 2, f"Expected 2 sleep calls, got {mock_sleep.call_count}"
+                    sleep_calls = [call[0][0] for call in mock_sleep.call_args_list]
+                    assert sleep_calls[0] == 1, f"First backoff should be 1s, got {sleep_calls[0]}"
+                    assert sleep_calls[1] == 2, f"Second backoff should be 2s, got {sleep_calls[1]}"
+
+    @pytest.mark.asyncio
+    async def test_recovery_mode_prompt_on_retry(self, executor):
+        """Test that retry uses recovery mode prompt with error context from previous attempt."""
+        retry_contexts = []
+        error_message = "Simulated error for testing"
+
+        async def capture_execute(subtask, attempt=1, recovery_context=None):
+            retry_contexts.append(recovery_context)
+            if attempt < 2:
+                return {"success": False, "error": error_message}
+            return {"success": True}
+
+        with patch.object(executor, "_execute_subtask_impl", side_effect=capture_execute):
+            subtask = {"id": "1", "title": "Test", "status": "pending"}
+            await executor.execute_subtask_with_retry(subtask)
+
+        # Verify recovery context structure
+        assert len(retry_contexts) == 2, f"Expected 2 attempts, got {len(retry_contexts)}"
+        assert retry_contexts[0] is None, "First attempt should have no recovery context"
+        assert retry_contexts[1] is not None, "Second attempt must have recovery context"
+        assert error_message in retry_contexts[1], (
+            f"Recovery context should contain previous error. Got: {retry_contexts[1]}"
+        )
+
+
+class TestCodingPhaseEscalation:
+    """Test escalation handling for persistent failures (Story 4.3 Task 6)."""
+
+    @pytest.mark.asyncio
+    async def test_escalate_after_max_retries(self, executor):
+        """Test that task is escalated after 3 failed retries."""
+
+        async def always_fail(subtask, attempt=1, recovery_context=None):
+            return {"success": False, "error": "Always fails"}
+
+        with patch.object(executor, "_execute_subtask_impl", side_effect=always_fail):
+            subtask = {"id": "1", "title": "Test", "status": "pending"}
+            result = await executor.execute_subtask_with_retry(subtask, max_retries=3)
+
+        assert result["success"] is False
+        assert result.get("escalated") is True or "escalated" in str(result.get("error", "")).lower()
+
+    @pytest.mark.asyncio
+    async def test_escalation_updates_task_state(self, executor):
+        """Test that escalation updates task state to 'escalated'."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            executor._task_dir = Path(tmpdir)
+
+            async def always_fail(subtask, attempt=1, recovery_context=None):
+                return {"success": False, "error": "Always fails"}
+
+            with patch.object(executor, "_execute_subtask_impl", side_effect=always_fail):
+                subtask = {"id": "1", "title": "Test", "status": "pending"}
+                await executor.execute_subtask_with_retry(subtask, max_retries=3)
+
+            from apps.backend.core.executors.full_auto import TaskState
+
+            state = executor.get_task_state()
+            assert state in [TaskState.ESCALATED.value, TaskState.FAILED.value, "escalated", "failed"]
+
+    @pytest.mark.asyncio
+    async def test_partial_progress_preserved_on_escalation(self, executor):
+        """Test that partial progress is preserved when task is escalated."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            spec_dir = Path(tmpdir)
+            plan_data = {
+                "subtasks": [
+                    {"id": "1", "title": "First", "status": "pending"},
+                    {"id": "2", "title": "Second", "status": "pending"},
+                ],
+            }
+            import json
+
+            plan_file = spec_dir / "implementation_plan.json"
+            plan_file.write_text(json.dumps(plan_data))
+            executor._spec_dir = spec_dir
+            executor._task_dir = Path(tmpdir)
+
+            call_count = 0
+
+            async def first_success_then_escalate(subtask, attempt=1, recovery_context=None):
+                nonlocal call_count
+                call_count += 1
+                if subtask.get("id") == "1":
+                    return {"success": True}
+                # Second subtask fails and escalates
+                return {"success": False, "error": "Always fails", "escalated": True}
+
+            with patch.object(executor, "_execute_subtask_impl", side_effect=first_success_then_escalate):
+                result = await executor.execute_coding_phase()
+
+            # Verify escalation occurred
+            assert result["status"] == "escalated"
+            assert result["subtask"] == "2"
+            assert result["subtasks_executed"] == 1
+
+            # Verify first subtask is still completed (preserved)
+            updated_plan = json.loads(plan_file.read_text())
+            assert updated_plan["subtasks"][0]["status"] == "completed"
+            # Second subtask should be marked as failed
+            assert updated_plan["subtasks"][1]["status"] == "failed"
+
+    @pytest.mark.asyncio
+    async def test_non_escalated_failure_stops_execution(self, executor):
+        """Test that a non-escalated failure stops execution and marks task failed.
+
+        This tests the edge case where execute_subtask_with_retry returns
+        {"success": False} without the "escalated" key.
+        """
+        with tempfile.TemporaryDirectory() as tmpdir:
+            spec_dir = Path(tmpdir)
+            plan_data = {
+                "subtasks": [
+                    {"id": "1", "title": "First", "status": "pending"},
+                    {"id": "2", "title": "Second", "status": "pending"},
+                ],
+            }
+            import json
+
+            plan_file = spec_dir / "implementation_plan.json"
+            plan_file.write_text(json.dumps(plan_data))
+            executor._spec_dir = spec_dir
+            executor._task_dir = Path(tmpdir)
+
+            # Simulate a failure that doesn't trigger escalation
+            async def non_escalated_fail(subtask, attempt=1, recovery_context=None):
+                return {"success": False, "error": "Network timeout"}
+
+            with patch.object(executor, "_execute_subtask_impl", side_effect=non_escalated_fail):
+                # Also patch execute_subtask_with_retry to return non-escalated failure
+                with patch.object(
+                    executor,
+                    "execute_subtask_with_retry",
+                    return_value={"success": False, "error": "Non-escalated failure"},
+                ):
+                    result = await executor.execute_coding_phase()
+
+            # Should fail, not silently continue
+            assert result["status"] == "failed"
+            assert "error" in result
+            from apps.backend.core.executors.full_auto import TaskState
+
+            state = executor.get_task_state()
+            assert state == TaskState.FAILED.value
+
+
+class TestCodingPhaseCompletion:
+    """Test state update on completion (Story 4.3 Task 7)."""
+
+    @pytest.mark.asyncio
+    async def test_coding_complete_state_on_success(self, executor):
+        """Test that successful coding phase updates state to coding_complete."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            executor._task_dir = Path(tmpdir)
+            executor._spec_dir = Path(tmpdir)
+
+            plan_data = {
+                "subtasks": [
+                    {"id": "1", "title": "Only subtask", "status": "pending"},
+                ],
+            }
+            import json
+
+            (executor._spec_dir / "implementation_plan.json").write_text(json.dumps(plan_data))
+
+            async def mock_execute(subtask, attempt=1, recovery_context=None):
+                return {"success": True}
+
+            with patch.object(executor, "_execute_subtask_impl", side_effect=mock_execute):
+                result = await executor.execute_coding_phase()
+
+            from apps.backend.core.executors.full_auto import TaskState
+
+            state = executor.get_task_state()
+            assert state == TaskState.CODING_COMPLETE.value or result.get("status") == "coding_complete"
+
+    @pytest.mark.asyncio
+    async def test_validation_triggered_after_coding_complete(self, mock_context):
+        """Test that validation phase is triggered after coding completes."""
+        phases = [
+            Phase(id="planning", name="Planning", order=0),
+            Phase(id="coding", name="Coding", order=1),
+            Phase(id="validation", name="Validation", order=2),
+        ]
+        runner = MockMethodologyRunner(phases=phases)
+        executor = FullAutoExecutor(
+            runner=runner,
+            context=mock_context,
+            task_config=mock_context.task_config,
+        )
+
+        await executor.execute()
+
+        # After coding, validation should automatically run
+        assert "coding" in runner._executed_phases
+        assert "validation" in runner._executed_phases
+
+    @pytest.mark.asyncio
+    async def test_completion_logged(self, executor):
+        """Test that coding phase completion is logged."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            executor._task_dir = Path(tmpdir)
+            executor._log_dir = Path(tmpdir)
+            executor._spec_dir = Path(tmpdir)
+
+            plan_data = {"subtasks": []}
+            import json
+
+            (executor._spec_dir / "implementation_plan.json").write_text(json.dumps(plan_data))
+
+            with patch.object(executor._logger, "info") as mock_info:
+                await executor.execute_coding_phase()
+
+                # Should have logged completion
+                calls_str = str(mock_info.call_args_list)
+                assert "complete" in calls_str.lower() or "finished" in calls_str.lower() or len(mock_info.call_args_list) > 0