How to modify/bypass a step that is returning an error?

[tacotuesday]

Question

Auto-GPT keeps trying to scrape a website that is returning a 403 error. This creates a loop where Auto-GPT summarizes the page and produces an empty text file that it cannot read. How do I change the source page or force Auto-GPT to print the summary in the terminal?

Background

I asked Auto-GPT to document the process and steps to build a new house in a specific location in the United States. Auto-GPT's response was fantastic:

Role:  an expert AI in home architecture and construction that assists individuals in documenting the process and steps required to build a new house in Saint Mary's County, Maryland, USA.
Goals: 
-  Provide a comprehensive and detailed guide on the step-by-step process of building a new house in Saint Mary's County, Maryland, USA, including obtaining permits, hiring contractors, and managing the construction timeline.
-  Offer expert advice on selecting suitable building materials, ensuring compliance with local building codes and regulations, and optimizing energy efficiency in the construction process.
-  Assist in creating a realistic budget and cost estimation for the entire construction project, including material costs, labor expenses, and potential unforeseen expenses.
-  Identify potential challenges and risks in the construction process and provide proactive solutions to mitigate them, ensuring a smooth and successful completion of the project.
-  Continuously update and improve the documentation based on the latest building codes, regulations, and industry best practices to ensure the information provided is accurate and up to date.
  Could not load MemoryItems from file: input length is 0: line 1 column 1 (char 0)
Using memory of type:  JSONFileMemory
Using Browser:  firefox
HARBPGPT THOUGHTS:  I should start by providing an overview of the step-by-step process of building a new house in Saint Mary's County, Maryland, USA.
REASONING:  This will help the user understand the overall process before diving into specific details.
PLAN: 
-  Provide an overview of the step-by-step process
-  Include information on obtaining permits, hiring contractors, and managing the construction timeline
-  Offer advice on selecting building materials and ensuring compliance with local regulations
-  Assist in creating a realistic budget and cost estimation
-  Identify potential challenges and provide solutions
-  Continuously update and improve the documentation
CRITICISM:  None
SPEAK:  I will provide an overview of the step-by-step process of building a new house in Saint Mary's County, Maryland, USA.

It runs web_search and returns several results. Auto-GPT summarizes one of the returned links, then writes a Python script to scrape the results:

import requests
from bs4 import BeautifulSoup

# Send a GET request to the website
response = requests.get('https://www.newhomesource.com/learn/step-by-step-guide-to-home-building-process/')

# Parse the HTML content
soup = BeautifulSoup(response.content, 'html.parser')

# Extract the relevant information
steps = soup.find_all('h2', class_='entry-title')

# Save the extracted information in a file
with open('building_steps.txt', 'w') as file:
    for step in steps:
        file.write(step.text + '\n')

This results in a 403 Forbidden error, so Auto-GPT tries to read_file an empty file. There appears to be no error handling, so Auto-GPT starts looping through the summarize-scrape-error loop until I manually break it. How do I get Auto-GPT to select the next relevant website?

[tacotuesday]

Error is on my side of the keyboard. I ran y -20 and didn't wait to see that I could provide input on the results.