Task: {{ task }}

Output Format:
```json
{
    "Observation": str,
    "Reasoning": str,
    "Current Step in Guidance Trajectory": "(int, str)",
    "Action": str,
    "Expectation": str
}
```

One Output Example:
```json
{
    "Observation": "The screenshot shows the google homepage of the Chrome browser.",
    "Reasoning": "The user wants to buy a lap-top on Amazon.com, so I need to click on the address bar of Chrome for entering the 'Amazon.com'.",
    "Current Step in Guidance Trajectory": "(2, 'Step 1 is completed in Action History, I am following the Guidance Trajectory step 2 to click on the address bar.')",
    "Action": "Click the search bar first to perform Google search and navigation",
    "Expectation": "The search button is activated after being clicked, ready to input."
}
```

Special notes for some computer-use actions:
{% include 'planner/special_action_note.txt' %}

Guidance Trajectory Examples:
{{ guidance_trajectory_example }}.

Action History (last {{ max_history_length }} steps) for current task execution:
{{ action_history_str }}.

**IMPORTANT NOTES:**
1. Carefully observe the screenshot, and read the guidance trajectories and history actions.
2. The Guidance Trajectory is a promise successful example of how to achieve the task. You should strictly follow the steps in the Guidance Trajectory unless you encounter unexpected issues. 
3. You can try to get the environment back to the state in the Trajectory.
4. Start from step 1 and complete the task step by step. DO NOT skip or overlook any step in the Guidance Trajectory.
5. You should only give one planning (one action) at a time.
6. Carefully determine the current step by comparing observation, action history with the Guidance Trajectory (e.g., if the last action of step n failed, you remain on step n, else move to step n+1). Only skip the steps (e.g., move to step n+2) if you are very sure the intermediate steps are completed.
7. When the task is completed, set "Action": null.

Now start planning:
