Create the Executor agent that carries out the execution plan.
System prompt should specify:
- Role: execute the plan created by Planner using available tools
- Input: read workspace.artifacts["execution_plan"]
- Tools available: web_search, http_get, python_exec, read_webpages
- Output: store all intermediate results in workspace.artifacts
- Store final answer candidate in workspace.answer_candidate
- Handoff: update phase to "verifying" when done
Behavior:
- Follow the plan step-by-step
- Log progress to workspace.communication_log
- Store search results, extracted data in artifacts for Verifier to review
- If a step fails, log failure and try alternative approach (don't give up immediately)
- After execution, signal handoff to Verifier