Summarize CLI

Problem

When summarizing web pages, local files, or YouTube content for indexing, comparison, or downstream automation, the hard part is often inconsistent input formats, anti-scraping limits, unreliable transcript retrieval, and uneven output length. Summarize-1.0 puts these sources behind one CLI: generate model summaries and emit machine-readable results with --json, which fits scripts, ingestion pipelines, and knowledge-management workflows.

How It Works

• Inputs: supports URLs, local files, and YouTube links, making it useful for extracting key points from heterogeneous sources.
• Model access: set OPENAI_API_KEY, ANTHROPIC_API_KEY, XAI_API_KEY, or GEMINI_API_KEY for your provider; the default model is google/gemini-3-flash-preview.
• Length control: use --length short|medium|long|xl|xxl and --max-output-tokens to tune summary granularity.
• Extraction and fallback: --extract-only handles URL content only; --firecrawl auto|off|always can use FIRECRAWL_API_KEY for restricted sites; --youtube auto enables Apify fallback when APIFY_API_TOKEN is configured.
• Output format: --json makes results easier to parse, store, or compare in batch.

Boundaries and Notes

This skill focuses on summary generation rather than full crawling or long-term memory. Highly dynamic pages, login-gated content, or paywalled material may still depend on external extraction services, and YouTube fallback depends on token and third-party availability. If downstream use requires auditing, keep the source URL, capture time, or local file path alongside the summary.