When should I put information in CLAUDE.md?

Information should only go in CLAUDE.md if it is a universal rule or guardrail needed most of the time to prevent mistakes during assistance.

Where are these Claude plugins and skills stored?

They are stored within the user's dotfiles directory, specifically under ~/.dotfiles/claude-plugins/ organized by category.

How does this skill handle AI mistakes?

It uses an escalation ladder: start by improving a skill, then add a project-specific hot rule if the error recurs, and finally move to a root CLAUDE.md guardrail if it persists across repos.

What is the difference between a Claude skill and a command?

Skills are auto-loaded by Claude based on context and descriptions, while commands are actions explicitly invoked by the user using a slash command.

Claude Memory Placement

Name: Claude Memory Placement
Author: mshuffett

bymshuffett

生产力与工作流

Optimizes AI knowledge retention by routing learnings, rules, and documentation to the most effective storage locations within the Claude environment.

关于

This skill provides a systematic framework for Claude to decide where information should be stored across sessions to ensure maximum efficiency and minimal error. It differentiates between universal guardrails in CLAUDE.md, context-aware skills for domain knowledge, and user-invoked commands for specific workflows. By following a structured decision tree and an escalation ladder for recurring mistakes, it helps maintain a clean and effective plugin-based dotfile architecture while ensuring critical instructions are available exactly when needed.

主要功能

Standardized plugin architecture for skills and commands
Best practices for writing trigger-based skill descriptions
Guidelines for maintaining project-specific vs. universal rules
Escalation ladder for managing and correcting recurring AI mistakes
0 GitHub stars
Structured decision tree for information routing

使用场景

Correcting repeated AI errors by progressively moving information from skills to hot rules
Organizing new development scripts into the correct plugin category (dev, media, terminal)
Determining if a new workflow rule belongs in a global or project-specific CLAUDE.md

关于

主要功能

Standardized plugin architecture for skills and commands
Best practices for writing trigger-based skill descriptions
Guidelines for maintaining project-specific vs. universal rules
Escalation ladder for managing and correcting recurring AI mistakes
0 GitHub stars
Structured decision tree for information routing

使用场景

Correcting repeated AI errors by progressively moving information from skills to hot rules
Organizing new development scripts into the correct plugin category (dev, media, terminal)
Determining if a new workflow rule belongs in a global or project-specific CLAUDE.md