Mimi Start — Development skill for Claude Code
Set a project up with mimi, once.
How to install Mimi Start
Installs to ~/.claude/skills/candybutcher27-memory-layer-mimi-start/SKILL.md
mkdir -p ~/.claude/skills/candybutcher27-memory-layer-mimi-start && curl -fsSL https://raw.githubusercontent.com/CandyButcher27/memory-layer/HEAD/commands/mimi-start.md -o ~/.claude/skills/candybutcher27-memory-layer-mimi-start/SKILL.md Restart Claude Code, or start a new session, for it to be picked up.
What Mimi Start does
description: Set a project up with mimi, once. Run again later only when you want a brief of where the last session stopped; the memory loads into every session without it. allowed-tools: Read, Edit, Write, Grep, Glob, Bash
/mimi-start
Get the project's memory ready. Follow the steps in order.
0. Find the script
- If
mimi/MIMI.mdexists, the script's path is on itsFind rot:line. - Otherwise use
$HOME/.claude/skills/mimi/memlayer.py. - If the file does not exist, stop and
Alternatives in Development
- Again — PUA Again — 用于再试试、换个方法、别放弃、为什么还不行、同一路径失败多次;要求停止微调同一思路,改用本质不同方案 19.5k ★
- Claude Supermemory — Persistent memory across sessions and projects using Supermemory 2.3k ★
- Working Memory — This is your working memory 192 ★
Full documentation available on GitHub
View Source RepositoryRelated Skills
Brief Me
Get briefed — loads your full memory, pulls live state from all connected tools, surfaces risks, staleness, up
Eli5
Say the last answer again in plain words — same facts, in the order the operator needs them. Optional topic ar
Thumbgate Guard
Turn the last agent mistake into a hard prevention rule the agent cannot bypass. Use after a bad tool call, a
Repurpose Skill
See a reel with a fire hook? Don't start from scratch. Drop the link, and Claude downloads it, transcribes it,
Mimi Close
Close a work session by writing what this session learned into the project's mimi/ folder (ISSUES, decisions,
Mimi Logging
Show what mimi costs and saves in this project, from its Claude Code session logs, and save the report under m
Related Agents
08 GitHub Ops
Git and GitHub workflow specialist. Invoked TWICE per brief by architect — FIRST at the start of the brief to
Save Memory
Step agent that executes memory writes from the routing manifest. Loads save specs once per run, processes eac
Context Scouter
Loads accumulated project memory (learnings, feedback, project facts) written by Context Keeper and returns a