summarize-chat

madrus's avatarfrom madrus

Make a summary of the chat - key decisions made, code patterns established, next steps identified. This saves a ton of tokens and thus AI-costs.

⭐0stars🔀0forks📁View on GitHub🕐Updated Jan 10, 2026

When & Why to Use This Skill

This Claude skill provides a structured framework for summarizing long chat sessions into high-density context prompts. By capturing critical technical decisions, architectural patterns, and pending tasks, it enables users to maintain project continuity while significantly reducing token usage and AI operational costs. It is an essential tool for developers and power users looking to optimize their context window and improve the efficiency of long-term AI interactions.

Use Cases

  • Cost-Effective Context Management: Condense lengthy coding conversations into a concise system prompt to save tokens and stay within model context limits during extended sessions.
  • Development Session Handoff: Generate a technical snapshot of work-in-progress, including modified files and established code patterns, to resume work seamlessly in a new chat or with a different agent.
  • Technical Decision Logging: Automatically document the rationale behind specific design choices, refactoring efforts, and architectural changes for project tracking and documentation.
  • Task Tracking & Next Steps: Extract and organize explicit and implicit pending tasks from a complex discussion to ensure no action items are missed in the development lifecycle.
namesummarize-chat
descriptionMake a summary of the chat - key decisions made, code patterns established, next steps identified. This saves a ton of tokens and thus AI-costs.

Chat Summary Skill

Create a concise summary (<=300 words) of our conversation to use as context for the next chat session.

Format Requirements

Structure the summary as a system prompt with these sections:

1. Context

  • What we were working on
  • Why this work matters to the project

2. Key Decisions Made

  • Technical decisions and their rationale
  • Design choices (patterns, colors, components)
  • Architecture changes

3. Code Patterns Established

  • Component patterns used
  • Styling approaches (variants, utilities)
  • Important conventions followed

4. Files Modified

  • List files changed with brief description of changes
  • Include specific line numbers if relevant
  • Note any refactoring (e.g., Panel → custom variants)

5. Technical Context

  • Important types, utilities, or APIs referenced
  • Relevant libraries or tools used
  • Key configuration or settings

6. Next Steps / Pending Tasks

  • Explicit tasks still to do
  • Implicit follow-up work identified
  • Testing or validation needed

Output Style

  • Be concise and technical
  • Focus on information needed to continue work seamlessly
  • Use bullet points for clarity
  • Include file paths and line numbers for quick reference
  • Emphasize decisions and rationale, not just what changed
  • Format as a system message that provides context without requiring re-reading the entire chat
  • Include word count at the end in format: [Word count: X/300]

Example Structure

Previous session context: We standardized title colors across the competition UI and implemented semi-transparent group panels using sky gradients.

Key decisions:
- Chose text-title class for darker primary color in titles
- Used sky-500/15 via-5 opacity pattern matching confirmed pool
- Refactored from Panel component to groupCardVariants for consistency

Files modified:
- CompetitionGroupsTab.tsx:29,52 - Updated titles to text-title
- GroupCard.tsx - Replaced Panel with section + groupCardVariants
- groupAssignment.variants.ts:37-70 - Sky gradient with dark mode

Current state: All changes implemented and passing typecheck and lint. Group panels now semi-transparent with sky color matching design system.

[Word count: 93/300]