fabric-patterns_skill

Overview

This skill runs danielmiessler/fabric CLI patterns to process content through curated AI prompts for analysis, extraction, summarization, writing, and security tasks. It selects an appropriate pattern, pipes content into fabric, and returns structured results integrated back into the conversation. Use it when pattern-driven, one-shot processing is more effective than inline responses.

How this skill works

I pick the most relevant fabric pattern based on your request, optionally run fabric's own suggest_pattern meta-pattern to confirm, and then execute the chosen pattern with the appropriate piping options. Inputs can be raw text, files, URLs, YouTube links, or transcripts; outputs can be streamed, saved, or returned inline. I summarize long outputs and offer the full pattern result on demand.

When to use it

• You want structured extraction from long articles, transcripts, or papers.
• You ask explicitly to "use fabric", "run a pattern", or name a specific pattern.
• You need YouTube or podcast summarization, chapters, or timestamped insights.
• You need security-focused analysis: threat models, incident reports, or Sigma rules.
• You need to chain patterns (e.g., extract_wisdom | summarize) for layered processing.

Best practices

• I will show the exact fabric command I intend to run and why before executing.
• Stream long outputs (fabric -s) to avoid truncation and improve readability.
• For large inputs, write to a temp file and pipe via cat rather than huge echoes.
• Respect pattern output formats; only reformat if you explicitly ask.
• When uncertain, I run fabric's suggest_pattern and compare suggestions before proceeding.

Example use cases

• Summarize a 5,000-word investigative article and extract key quotes and takeaways.
• Analyze a threat report and produce indicators, TTPs, and suggested detections.
• Generate video chapters and a concise timestamped summary from a YouTube lecture.
• Run extract_wisdom on a podcast transcript and produce highlighted insights for marketing.
• Chain extract_insights then summarize to create a 20-word hook plus three bullets for social.