Skip to main content
skill CX Codex

summarization — Claude Code skill: what it does, how to install

Selects the right summary format based on audience, goal, and length. Turns raw text into executive summaries, TLDRs, or research abstracts.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add jwynia/agent-skills --skill summarization -a claude-code
Open on GitHub

What summarization does

This skill treats summarization as translation rather than mechanical text reduction. It directs the agent to clarify five aspects before condensing content: purpose, audience, scope, emphasis, and format. That stops shallow data dumps.

The skill defines concrete reduction models. Business documents get an executive summary highlighting decisions and next steps. Research material gets structured abstracts or critical reviews. Brief tasks condense into a crisp TLDR or key point extraction.

Without a defined purpose, AI summaries often smooth over vital details. The built-in quality checklist prevents this by watching for meaning distortion and excessive abstraction.

What it helps with

  • draft an executive summary covering decisions and next steps
  • synthesize multiple research sources into a unified narrative
  • distill long documents into a one-paragraph TLDR
  • extract discrete claims and verbatim quotes from source text

Example prompts for the agent

  • Generate an executive summary for project-proposal.md focusing on risks and actions
  • Create a one-sentence TLDR from customer-feedback.txt
  • Synthesize findings across the reports in the research folder and point out contradictions

How to install summarization

1Skills CLI (Vercel)

npx skills add jwynia/agent-skills --skill summarization -a claude-code

2Manually: put the files in place

~/.claude/skills/summarization/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/summarization/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about summarization

How to install summarization in Claude Code?

Run npx skills add jwynia/agent-skills --skill summarization -a claude-code. The file will be placed in ~/.claude/skills/summarization/SKILL.md.

What does the summarization skill do?

It guides agents to select a summarization approach matching the audience and context. It formats output into TLDRs, outlines, abstracts, or executive summaries instead of standard reductions.

Which anti-patterns does this skill address?

It prevents four common mistakes: dumping raw facts without reducing complexity, distorting the original argument, applying one format to all requests, and over-abstracting useful specifics.

Source text

SKILL.md · MIT
---
name: summarization
description: "Create effective summaries by matching summarization type to purpose, audience, and context. Use when asked to summarize, create TLDR, condense content, or create executive summaries. Keywords: summary, TLDR, condense, executive summary, abstract."
license: MIT
metadata:
  author: jwynia
  version: "1.0"
  type: utility
  mode: generative
  domain: writing
---

# Summarization

## Purpose

Create effective summaries by matching summarization type to purpose, audience, and context. "Summarize" can mean many different things—this skill helps identify and execute the right approach.

## Core Principle

**Summarization is translation, not just reduction.** Different purposes require different summary types. Clarify the need before condensing.

---

## Clarifying Questions

Before summarizing, consider:

1. **Purpose:** What will this summary be used for?
   - Decision-making
   - Background information
   - Further research
   - Quick understanding
   - Reference/recall

2. **Audience:** Who will read it?
   - Technical experts
   - General audience
   - Decision makers
   - Familiar/unfamiliar with topic

3. **Scope:** How comprehensive?
   - Ultra-brief (single sentence)
   - Brief (paragraph)
   - Moderate (page)
   - Extended (multiple pages)

4. **Emphasis:** What aspects are most important?
   - Methodology
   - Findings/results
   - Arguments/claims
   - Context/background
   - Implications/applications

5. **Format:** What structure?
   - Narrative text
   - Bullet points
   - Hierarchical outline
   - Visual representation

---

## Summary Type Taxonomy

### Information Reduction Approaches

| Type | What It Is | When to Use |
|------|-----------|-------------|
| **Key Point Extraction** | Isolating the most important claims | Original has discrete important points |
| **Abstraction** | Higher-level statements covering multiple details | Patterns matter more than specifics |
| **Gisting** | Capturing essential meaning, discarding details | Only core message matters |
| **Compression** | Shortening while preserving information | Comprehensive coverage needed in less space |

### Structural Approaches

| Type | What It Is | When to Use |
|------|-----------|-------------|
| **Executive Summary** | Business-focused: decisions, recommendations, outcomes | Documents requiring action |
| **Abstract/Précis** | Academic: methodology and findings | Research papers, technical documents |
| **TLDR** | Ultra-brief main takeaway | Casual communication, extreme brevity |
| **Outline** | Hierarchical structure of main/supporting points | Logical structure matters |

### Purpose-Oriented Approaches

| Type | What It Is | When to Use |
|------|-----------|-------------|
| **Synthesis** | Combining multiple sources coherently | Summarizing across documents |
| **Critical Summary** | Evaluating claims while condensing | Assessment of quality needed |
| **Contextual Summary** | Framing within broader knowledge | Understanding bigger picture matters |
| **Actionable Summary** | Focusing on implications and next steps | Summary will drive action |

---

## Execution by Type

### Key Point Extraction
1. Scan for topic sentences and conclusions
2. Identify explicitly stated main ideas
3. List each distinct point
4. Preserve original phrasing where powerful

**Example:** "The author makes three main arguments: (1)..., (2)..., (3)..."

### Abstraction
1. Group related details
2. Find common themes or patterns
3. Create higher-level statements
4. Reduce specifics to principles

**Example:** "Multiple studies consistently show..." instead of listing 12 studies

### Gisting
1. Ask: "What is the one thing to remember?"
2. Distill to core insight
3. Remove all supporting detail
4. Verify essence is preserved

**Example:** "Remote work increases productivity for most knowledge workers."

### Executive Summary
1. State the problem/opportunity
2. Present the solution/recommendation
3. Highlight key benefits
4. Note costs/risks
5. Specify required actions

### Synthesis
1. Read all sources
2. Identify common themes
3. Note contradictions
4. Find complementary information
5. Create unified narrative

**Example:** "Across the five reports, three key trends emerge..."

### Critical Summary
1. Summarize the claims
2. Evaluate the evidence
3. Assess methodology
4. Note limitations
5. Conclude with assessment

**Example:** "While the author claims X, the evidence is limited by..."

---

## Format Variations

### Quotation-Based
- Use key phrases verbatim
- When precise wording is important
- Select and organize most important quotes

### Bullet Points
- Break continuous text into discrete units
- When quick scanning is valued
- Make each point standalone

### Progressive Summary
- Start ultra-brief
- Add layers of detail
- Let reader choose depth

### Comparative Summary
- Side-by-side analysis
- Highlight similarities and differences
- When contrasting sources

---

## Quality Checklist

- [ ] Purpose is clear
- [ ] Audience is considered
- [ ] Scope is appropriate
- [ ] Emphasis matches needs
- [ ] Format serves purpose
- [ ] Core message is preserved
- [ ] Reduction is proportional
- [ ] No invented information
- [ ] Attribution where needed

---

## Anti-Patterns

### The Information Dump
**Problem:** Reduces length but not complexity
**Fix:** Focus on what matters, not just what's short

### The Distortion
**Problem:** Changes meaning through compression
**Fix:** Verify summary against original claims

### The One-Size-Fits-All
**Problem:** Same approach for all requests
**Fix:** Match type to purpose and audience

### The Over-Abstraction
**Problem:** Loses all useful specifics
**Fix:** Preserve concrete details that support understanding

---

## Integration Points

**Inbound:**
- When asked to summarize any content
- When processing long documents
- When creating documentation

**Outbound:**
- To decision-making processes
- To knowledge management systems
- To communication outputs

**Complementary:**
- `speech-adaptation`: For spoken summaries
- `voice-analysis`: For maintaining voice in summaries

© jwynia · MIT · jwynia/agent-skills

Similar skills

Everything for Writing and docs →

From the same set81

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor