summary-generator
Generates session summaries and resume prompts for multi-session work. Use when completing features, before context limits (~50% capacity), or when user says "summary", "wrap up", "save progress", "end session". Creates markdown in docs/summaries/ with completed work, files modified, and restart instructions.
When & Why to Use This Skill
This Claude skill automates the creation of detailed session summaries and resume prompts, effectively solving the problem of context loss and productivity decay in long-running development tasks. By documenting completed work, git modifications, and pending tasks into standardized markdown files, it enables seamless transitions between AI sessions and ensures the agent maintains a high-fidelity 'memory' even when hitting context window limits.
Use Cases
- Managing complex feature development where the conversation history nears context limits, requiring a fresh session without losing project state.
- Generating structured hand-off documentation for team collaboration to explain recent architectural decisions and file changes.
- Maintaining an automated project changelog in docs/summaries/ to track daily progress and git diffs systematically.
- Creating 'Resume Prompts' that allow a user to instantly re-sync the AI agent in a new chat with all necessary background and immediate next steps.
| name | summary-generator |
|---|---|
| description | Generates session summaries and resume prompts for multi-session work. Use when completing features, before context limits (~50% capacity), or when user says "summary", "wrap up", "save progress", "end session". Creates markdown in docs/summaries/ with completed work, files modified, and restart instructions. |
| allowed-tools | Read, Glob, Grep, Bash(git diff:*), Bash(git log:*), Bash(git status:*), Write |
Session Summary Generator
Overview
This skill creates comprehensive session summaries for complex multi-session work, enabling seamless resumption of tasks. It generates a markdown file in docs/summaries/ with a standardized format.
When to Use
Trigger this skill when:
- User explicitly requests a summary ("generate summary", "wrap up", "save progress")
- Completing a significant feature or refactor
- Conversation context is reaching limits (~50% before auto-compact)
- Before starting a new chat session
- When collaborating with team members on the same feature
Auto-Suggest Triggers
Proactively suggest generating a summary when:
- Multiple files have been modified in the session
- A feature implementation is complete
- The conversation has been long (many exchanges)
- User mentions ending their work session
Output Location
Session summaries are stored in: docs/summaries/YYYY-MM-DD_feature-name.md
Instructions
Step 1: Analyze Current Work
Run these commands to understand what was done:
git status
git diff --stat
git log --oneline -10
Review the conversation history to identify:
- What was accomplished
- Key decisions made
- Files created or modified
- Any remaining tasks
Step 2: Generate Summary File
Create the summary using the template in TEMPLATE.md.
Key sections to include:
- Overview: Brief description of session focus
- Completed Work: Bullet points of accomplishments
- Key Files Modified: Table of files and changes
- Design Patterns Used: Important architectural decisions
- Remaining Tasks: What's left to do
- Resume Prompt: Copy-paste instructions for next session
Step 3: Create Resume Prompt
The resume prompt should include:
- Context reference to the summary file
- Specific file paths to review first
- Current status and immediate next steps
- Any blockers or decisions that need user input
Example Usage
When user says: "Let's wrap up for today"
Response:
- Analyze git changes and conversation history
- Create
docs/summaries/2025-12-30_enrollment-improvements.md - Provide the resume prompt for the next session
- Suggest: "When context gets long, consider starting a new chat with the resume prompt"
Tips
- Keep summaries focused on a single feature or area
- Include exact file paths for easy navigation
- Note any environmental setup needed (database migrations, etc.)
- Flag any blocking issues or decisions made
- Reference the CLAUDE.md file patterns when applicable