Have Claude collect names from a Wikipedia page into a file
The poster tried Claude Computer Use to open Firefox, search Wikipedia, find all names on a page and save them to a text file. Task: Suggested implementation: Use Claude Computer Use to control Firefox, navigate to a Wikipedia page, extract all names, and save them to a text file. The workflow involves running Anthropic’s Docker image locally with an API key. Human role: Verify extracted names against the source page.; tools: Claude computer use. Human role: The original post does not say how a person stayed involved.; limitations: Prerequisites include a local Docker environment and an Anthropic API key. The primary constraint is high token cost due to screenshot processing. Source reliability is secondary, as the original Reddit link returned a 403 error. Human approval is required for cost validation..
Tools
Published on Agent Work Atlas:
Overview
The poster tried Claude Computer Use to open Firefox, search Wikipedia, find all names on a page and save them to a text file.
Workflow
Suggested implementation: Use Claude Computer Use to control Firefox, navigate to a Wikipedia page, extract all names, and save them to a text file. The workflow involves running Anthropic’s Docker image locally with an API key. Human role: Verify extracted names against the source page.
Reported result
The task reportedly completed in about one minute for simple extraction. However, it consumed nearly 100,000 tokens, costing approximately $0.31. The poster considered this cost too high for such a trivial task. No other failures were explicitly reported.
Human role
The original post does not say how a person stayed involved.
Limitations
Prerequisites include a local Docker environment and an Anthropic API key. The primary constraint is high token cost due to screenshot processing. Source reliability is secondary, as the original Reddit link returned a 403 error. Human approval is required for cost validation.
Try it
Act as a web automation agent. I need you to extract names from a specific Wikipedia page using Claude Computer Use. Inputs: - Target URL: [replaceable input: e.g., https://en.wikipedia.org/wiki/List_of_Nobel_laureates] - Output Path: [replaceable input: ./names.txt] Steps: 1. Launch Firefox via Claude Computer Use. 2. Navigate to the Target URL. 3. Identify and extract all visible names on the page. 4. Save the list to the Output Path. Safeguards: - Perform a dry run on a small section first. - Stop if token usage exceeds [replaceable input: 50,000] tokens. - Require human confirmation before finalizing the file. Deliverables: - A text file containing the names. - A report on token usage and cost. Quality Check: - Verify 10 random names against the source page. - Ensure no duplicates. If inputs are missing, ask for the Target URL and Output Path. Do not proceed without human confirmation on the dry run results.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.