lets-wrap

Capture session learnings and commit approved edits to mission-context and CLAUDE.md.

Updated Apr 10, 2026
One-click install
npx skills add https://github.com/GavinTPA/ai-powerhouse-starter --skill lets-wrap
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: lets-wrap
Source: https://github.com/GavinTPA/ai-powerhouse-starter/tree/main/my-skills/lets-wrap
Command: npx skills add https://github.com/GavinTPA/ai-powerhouse-starter --skill lets-wrap

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Lets-wrap prevents session learnings from being lost by capturing corrections, preferences, and discoveries from a conversation and encoding them into your workspace so future sessions start smarter.

Core Features & Use Cases

  • Session review: Summarises what was worked on today in plain English without exposing raw git output.
  • Learning capture & classification: Scans the conversation for direct corrections, patterns, and observations and classifies them as HIGH, MEDIUM, or LOW with coaching-style prompts.
  • Context updates & safe save: Proposes one-at-a-time edits to mission-context and CLAUDE.md, requests explicit approval via AskUserQuestion, and commits approved changes to GitHub silently.
  • Use case: After drafting a newsletter, run lets-wrap to record your preferred voice, save terminology changes to mission context, and back everything up.

Quick Start

Say lets wrap at the end of your session to capture learnings, review proposed edits, and save approved updates to your repository.

Frequently Asked Questions about lets-wrap

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I save Claude session learnings and user preferences to my repository?

You can capture session learnings by running an end-of-session workflow that scans the conversation for corrections and patterns, classifies them by importance, and commits approved edits directly to your mission-context and CLAUDE.md files via git.

What is the best way to update CLAUDE.md and mission-context files without exposing raw git output?

Updating CLAUDE.md and mission-context silently is possible by running background git add, commit, and push operations after proposing changes one at a time and receiving explicit user approval through interactive prompts.

How does AI memory classification work when capturing session discoveries?

AI memory classification works by scanning the conversation for direct corrections, patterns, and observations, then categorizing each finding as HIGH, MEDIUM, or LOW priority using coaching-style prompts before proposing any file edits.

Can I use interactive approval prompts to review proposed context updates before committing?

Yes, you can use AskUserQuestion interactive prompts to review proposed edits to your mission-context and CLAUDE.md files one at a time, ensuring you explicitly approve every change before it is pushed to your repository.

Do I need a pre-configured CLAUDE.md file to start capturing AI session context?

Yes, you need existing CLAUDE.md and mission-context file locations within your repository, as the workflow reads these locations to propose targeted edits and silently commit approved changes back to your workspace.

What are the limitations of automatically committing AI session learnings to a repository?

The primary limitation is that no changes are saved without explicit user approval, meaning you must manually review and accept each proposed edit, and the silent git workflow is restricted to repositories already containing mission-context and CLAUDE.md files.