How does a lesson differ from an observation?

Lessons are proactive rules that trigger specifically before a tool is used (PreToolUse), whereas observations are passive memories that Claude may or may not recall during a conversation.

What is the benefit of using trigger_pattern?

The trigger_pattern allows you to use regex to match specific command inputs, such as preventing a specific CLI flag or ensuring a certain directory is always included in a command.

Are lessons shared across different projects?

By default, lessons are scoped to a specific project path. However, you can create global lessons by omitting the project parameter during creation.

Can I apply lessons to specific tools only?

Yes, you can use the trigger_tool parameter to target specific actions like Bash, Edit, Write, or NotebookEdit to ensure rules only fire when relevant.

When should I use create_lesson?

You should use it immediately after correcting a mistake made by the AI to ensure the same error is caught proactively in future sessions.

Proactive Memory Lessons

Name: Proactive Memory Lessons
Author: metazen11

bymetazen11

0•

生産性とワークフロー

Establishes proactive rules and safety checks that trigger before risky AI operations to prevent recurring coding mistakes.

The create_lesson skill enables AI coding agents to learn from past errors by creating persistent, proactive instructions that activate during the PreToolUse phase. Unlike passive observations, these lessons are injected at the start of a session and triggered by specific tool interactions—such as Bash commands or file edits—ensuring the agent validates constraints before executing potentially dangerous actions. This skill is essential for building a project-specific guardrail system that improves agent reliability and safety over time.

主な機能

01Project-level scoping to ensure rules only apply to relevant codebases

02Persistent memory that carries 'lessons learned' across multiple sessions

03Proactive PreToolUse hooks to intercept risky actions before execution

04Tool-specific triggers for targeting Bash, Edit, Write, or Notebook tools

050 GitHub stars

06Regex pattern matching for fine-grained control over command execution

ユースケース

01Enforcing specific architectural patterns or linting rules before file edits

02Blocking known problematic command sequences that previously caused errors

03Preventing accidental production deployments by enforcing a config diff check

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add metazen11/agent-memory create_lesson

For use in Claude.ai and ChatGPT

Download Skill