summarize

What This Skill Does

The summarize skill is a robust, high-performance command-line utility designed to distill information from a wide variety of sources into concise, digestible formats. Whether you are dealing with sprawling webpages, dense PDF documents, high-resolution images, audio transcripts, or lengthy YouTube videos, this agent skill leverages state-of-the-art AI models to extract core concepts and key takeaways. It acts as an intelligent information filter, allowing users to move past information overload by focusing only on the data points that matter. By supporting multiple API providers including OpenAI, Anthropic, xAI, and Google, it offers unparalleled flexibility for users to select the underlying intelligence that best suits their cost or performance requirements.

Installation

To integrate this utility into your OpenClaw environment, ensure you have the OpenClaw CLI installed, then execute the following command in your terminal:

clawhub install openclaw/skills/skills/steipete/summarize

Once installed, verify that your environment variables (such as OPENAI_API_KEY or GEMINI_API_KEY) are correctly configured to allow the skill to authenticate with your preferred LLM providers. You may optionally create a configuration file at ~/.summarize/config.json to define a persistent default model, ensuring a consistent experience every time you invoke the tool.

Use Cases

This skill is ideal for researchers, students, and software developers who need to digest massive amounts of content quickly. For instance, you can use it to summarize technical documentation PDFs before starting a new project, extract key arguments from long-form investigative journalism articles, or generate concise bullet points from hours of YouTube educational content. It is particularly powerful for rapid content auditing, allowing you to scan dozens of sources for relevance before committing time to deep-reading.

Example Prompts

1. "Summarize the following technical paper at https://example.com/research-paper.pdf using a medium length output and format it as JSON."
2. "Watch the YouTube video at https://youtu.be/dQw4w9WgXcQ and provide an XL summary of the main arguments presented."
3. "Extract the core insights from my local document at ~/reports/annual_meeting.pdf using the Gemini 3 Flash model."

Tips & Limitations

For complex or anti-bot protected websites, be sure to set your FIRECRAWL_API_KEY to ensure the scraper can retrieve the content effectively. If you frequently summarize YouTube content, configuring the APIFY_API_TOKEN provides a reliable fallback for transcript extraction. Keep in mind that extremely long inputs may require adjusting the --max-output-tokens flag to avoid truncation. While highly capable, always verify the accuracy of the summary if the source material contains highly technical or esoteric data, as LLMs can sometimes hallucinate specific figures.