What kind of artifacts can it create?

It can generate Rules (declarative heuristics), Skills (executable command sequences), Hooks (automatic triggers), or Agent updates (enhanced personality/capabilities).

What is meta-pattern detection?

It is a process where the skill looks at clusters of learnings; if many rules relate to one topic, it suggests creating a dedicated Skill to handle that domain more efficiently.

How does Compound Learnings identify what to save?

It scans session logs for specific headers like 'Actionable Takeaways' and 'What Worked', then uses a frequency table to suggest artifacts only when a pattern appears across multiple sessions.

Where does the skill store its data?

It operates within the .maestro/ directory of your project, looking at cache/learnings and writing to the rules, skills, and hooks subdirectories.

Can I review changes before they are applied?

Yes, the skill is designed to present proposals including the rationale and draft content, requiring user approval via the AskUserQuestion tool before creation.

Compound Learnings

Name: Compound Learnings
Author: scooter-lacroix

byscooter-lacroix

생산성 및 워크플로

Transforms ephemeral session insights into permanent rules, skills, and automation hooks to continuously improve your development environment.

소개

Compound Learnings is a meta-capability for Claude Code that builds a persistent organizational memory for your development projects. By analyzing recent session logs, it identifies recurring patterns, success heuristics, and common pitfalls to suggest permanent improvements. Instead of losing insights when a session ends, this skill categorizes them into declarative rules, executable skills, or automated hooks. It utilizes a signal-based threshold system to ensure only high-value, recurring patterns are promoted, effectively allowing your AI development environment to grow smarter and more specialized with every task you complete.

주요 기능

0 GitHub stars
Meta-pattern detection to consolidate multiple rules into unified skills
Signal-based thresholding to identify high-confidence improvements
Categorization of insights into Rules, Skills, Hooks, or Agent updates
Automated extraction of patterns from recent session learnings
Structured proposal system for manual or batch approval of new artifacts

사용 사례

Improving project setup based on recurring debugging patterns
Codifying project-specific standards after a major feature implementation
Converting a manual multi-step workflow into a permanent executable skill

소개

주요 기능

0 GitHub stars
Meta-pattern detection to consolidate multiple rules into unified skills
Signal-based thresholding to identify high-confidence improvements
Categorization of insights into Rules, Skills, Hooks, or Agent updates
Automated extraction of patterns from recent session learnings
Structured proposal system for manual or batch approval of new artifacts

사용 사례

Improving project setup based on recurring debugging patterns
Codifying project-specific standards after a major feature implementation
Converting a manual multi-step workflow into a permanent executable skill