Lifting Loops

A retro skill that finds the loops where you keep acting as a manual step for your agents and proposes concrete ways to lift you out of them.

AutomationAdvanced★ 410⑂ 26AI score 9/10Last updated: Oct 29, 2026

What it does

Reviews a session (or recent local agent logs) and lists every human intervention, sorted into six roles: retry, judge, approve, carry context, trigger by hand, stand in for the agent.
Keeps a private loop ledger (private/ledger.md) with evidence and counts, and drafts a proposal (private/proposals/P-NNN.md) once a loop becomes a candidate.
Picks the lightest exit first: delete a stale rule → edit an existing AGENTS.md / CLAUDE.md / playbook → add a script, lint or metric → add a new skill → add a scheduled task or hook → global rule → advice about your own habits.
Applies accepted proposals on command ("accept P-003"), prunes redundant auto-memory files, and logs weekly intervention counts plus an outcome signal so "the user gave up" isn't mistaken for improvement.
Enforces redaction rules so nothing personal, internal, or absolute-path lands in git-tracked files.

Who it's for

Daily Claude Code / Codex users who notice they keep typing the same nudges.
People who actively maintain CLAUDE.md, skills, and scheduled tasks and want automation to actually improve.
Power users who want a weekly retro across session logs, memory files, and an Obsidian vault.

Examples

After a long task, say "reflect on this session" — you get a list of your interventions, ledger updates, and new proposal numbers with one-line summaries.
Run it in an empty session and it scans recent Claude Code / Codex / MiniMax sessions plus notes, returning a report with the loop most worth fixing this week.
Say "accept P-003" and it opens the target file (e.g. a project AGENTS.md), applies the change, deletes the memory files the proposal superseded, and marks the loop graduated.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/MiniMax-AI/mcode-stack/HEAD/stacks/jiachen/skills/lifting-loops/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)

Download the ZIP with the button below.
In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.

↓ Download ZIP

Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on AI Skill Mart.
Copy the stacks/jiachen/skills/lifting-loops folder from the GitHub repo MiniMax-AI/mcode-stack into my ~/.claude/skills/lifting-loops/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/MiniMax-AI/mcode-stack.git && mkdir -p ~/.claude/skills && cp -r mcode-stack/stacks/jiachen/skills/lifting-loops ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

Open a terminal and change into a working directory.
Clone the repo: git clone https://github.com/MiniMax-AI/mcode-stack.git
Create the skills directory: mkdir -p ~/.claude/skills
Copy the skill in: cp -r mcode-stack/stacks/jiachen/skills/lifting-loops ~/.claude/skills/
Verify Python 3 is available with python3 --version (batch mode runs scripts/digest.py).
Restart Claude Code, then invoke /lifting-loops or just ask it to "reflect on this session".
On first run it asks for your Obsidian vault path and claude.ai export path — leave them blank to scan only the default sources.
For the weekly automatic run, follow references/scheduling.md inside the skill folder.

View source on GitHub ↗License: MIT