feat(backend): Add Celery worker infrastructure with task stubs

- Add Celery app configuration with Redis broker/backend
- Add task modules: agent, workflow, cost, git, sync
- Add task stubs for:
  - Agent execution (spawn, heartbeat, terminate)
  - Workflow orchestration (start sprint, checkpoint, code review)
  - Cost tracking (record usage, calculate, generate report)
  - Git operations (clone, commit, push, sync)
  - External sync (import issues, export updates)
- Add task tests directory structure
- Configure for production-ready Celery setup

Implements #18

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

backend/app/tasks/agent.py

@@ -0,0 +1,150 @@
+# app/tasks/agent.py
+"""
+Agent execution tasks for Syndarix.
+
+These tasks handle the lifecycle of AI agent instances:
+- Spawning new agent instances from agent types
+- Executing agent steps (LLM calls, tool execution)
+- Terminating agent instances
+
+Tasks are routed to the 'agent' queue for dedicated processing.
+"""
+
+import logging
+from typing import Any
+
+from app.celery_app import celery_app
+
+logger = logging.getLogger(__name__)
+
+
+@celery_app.task(bind=True, name="app.tasks.agent.run_agent_step")
+def run_agent_step(
+    self,
+    agent_instance_id: str,
+    context: dict[str, Any],
+) -> dict[str, Any]:
+    """
+    Execute a single step of an agent's workflow.
+
+    This task performs one iteration of the agent loop:
+    1. Load agent instance state
+    2. Call LLM with context and available tools
+    3. Execute tool calls if any
+    4. Update agent state
+    5. Return result for next step or completion
+
+    Args:
+        agent_instance_id: UUID of the agent instance
+        context: Current execution context including:
+            - messages: Conversation history
+            - tools: Available tool definitions
+            - state: Agent state data
+            - metadata: Project/task metadata
+
+    Returns:
+        dict with status and agent_instance_id
+    """
+    logger.info(
+        f"Running agent step for instance {agent_instance_id} with context keys: {list(context.keys())}"
+    )
+
+    # TODO: Implement actual agent step execution
+    # This will involve:
+    # 1. Loading agent instance from database
+    # 2. Calling LLM provider (via litellm or anthropic SDK)
+    # 3. Processing tool calls through MCP servers
+    # 4. Updating agent state in database
+    # 5. Scheduling next step if needed
+
+    return {
+        "status": "pending",
+        "agent_instance_id": agent_instance_id,
+    }
+
+
+@celery_app.task(bind=True, name="app.tasks.agent.spawn_agent")
+def spawn_agent(
+    self,
+    agent_type_id: str,
+    project_id: str,
+    initial_context: dict[str, Any],
+) -> dict[str, Any]:
+    """
+    Spawn a new agent instance from an agent type.
+
+    This task creates a new agent instance:
+    1. Load agent type configuration (model, expertise, personality)
+    2. Create agent instance record in database
+    3. Initialize agent state with project context
+    4. Start first agent step
+
+    Args:
+        agent_type_id: UUID of the agent type template
+        project_id: UUID of the project this agent will work on
+        initial_context: Starting context including:
+            - goal: High-level objective
+            - constraints: Any limitations or requirements
+            - assigned_issues: Issues to work on
+            - autonomy_level: FULL_CONTROL, MILESTONE, or AUTONOMOUS
+
+    Returns:
+        dict with status, agent_type_id, and project_id
+    """
+    logger.info(
+        f"Spawning agent of type {agent_type_id} for project {project_id}"
+    )
+
+    # TODO: Implement agent spawning
+    # This will involve:
+    # 1. Loading agent type from database
+    # 2. Creating agent instance record
+    # 3. Setting up MCP tool access
+    # 4. Initializing agent state
+    # 5. Kicking off first step
+
+    return {
+        "status": "spawned",
+        "agent_type_id": agent_type_id,
+        "project_id": project_id,
+    }
+
+
+@celery_app.task(bind=True, name="app.tasks.agent.terminate_agent")
+def terminate_agent(
+    self,
+    agent_instance_id: str,
+    reason: str,
+) -> dict[str, Any]:
+    """
+    Terminate an agent instance.
+
+    This task gracefully shuts down an agent:
+    1. Mark agent instance as terminated
+    2. Save final state for audit
+    3. Release any held resources
+    4. Notify relevant subscribers
+
+    Args:
+        agent_instance_id: UUID of the agent instance
+        reason: Reason for termination (completion, error, manual, budget)
+
+    Returns:
+        dict with status and agent_instance_id
+    """
+    logger.info(
+        f"Terminating agent instance {agent_instance_id} with reason: {reason}"
+    )
+
+    # TODO: Implement agent termination
+    # This will involve:
+    # 1. Loading agent instance
+    # 2. Updating status to terminated
+    # 3. Saving termination reason
+    # 4. Cleaning up any pending tasks
+    # 5. Sending termination event
+
+    return {
+        "status": "terminated",
+        "agent_instance_id": agent_instance_id,
+    }