What output formats are available?

Users can request summaries in structured markdown, bullet points, TL;DR one-liners, JSON, tabular grids, or hierarchical outlines.

How does this skill prevent AI hallucinations?

It enforces strict fidelity rules including 'Read Before Summarizing' and 'Extract Before Abstracting,' ensuring all summary points are grounded in direct evidence from the source.

Can the Summarizer handle multiple sources at once?

Yes, for three or more sources, the skill can spawn a specialized agent team to parallelize the work and synthesize the findings into a single cohesive report.

Does it support image analysis?

Yes, the skill includes an image-summarization strategy that uses multimodal capabilities to describe visual content like screenshots, diagrams, and UI layouts.

What is included in the mandatory structured output?

Every summary includes YAML metadata, a BLUF-style summary, specific findings with references, a section for what was NOT found, and an explicit confidence level.

High-Fidelity Summarizer

Name: High-Fidelity Summarizer
Author: Jamie-BitFlight

byJamie-BitFlight

•

生产力与工作流

Standardizes and automates high-accuracy summarization across files, URLs, and images while enforcing strict anti-hallucination rules.

The Summarizer skill provides a robust framework for Claude to handle complex information extraction tasks with architectural precision. It intelligently routes requests based on source type—whether it's a local file, a web URL, or a visual image—and applies specific templates for formats like JSON, tables, or hierarchical outlines. By enforcing mandatory fidelity rules such as 'extract before abstracting' and 'preserving exact counts,' it eliminates common AI hallucinations and ensures that summaries are grounded, structured, and verifiable, making it an essential tool for developers and researchers managing multi-source data sets.

主要功能

0120 GitHub stars

02Strict anti-hallucination fidelity rules to ensure data accuracy

03Support for multiple output formats including JSON, tables, and outlines

04Automated team coordination for synthesizing 3+ sources simultaneously

05Mandatory structured output with YAML frontmatter and confidence levels

06Intelligent routing for files, URLs, images, and multi-source content

使用场景

01Distilling complex technical documentation and source code into structured highlights

02Generating consistent, high-fidelity descriptions of visual assets and screenshots

03Synthesizing findings from multiple research papers or web URLs with cross-referencing

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jamie-bitflight/claude_skills summarizer

For use in Claude.ai and ChatGPT

Download Skill