What makes a 'good' learning according to this skill?

A high-quality learning must be specific, testable, likely to recur, and grounded in actual friction or success documented during the session.

Can I prioritize specific topics during reflection?

Yes. If you provide freeform text with your reflection request, the skill treats it as a primary lens and validates it against the session data.

How do I trigger the reflection process?

You can invoke the reflection workflow by using the /compound:reflect command or by asking Claude to 'debrief this work' or 'capture learnings' from the current session.

What is the difference between frontier-shifting and frontier-sliding?

Frontier-shifting learnings improve underlying capabilities to increase both quality and volume of output, while frontier-sliding optimizes within existing constraints.

Where are the captured learnings stored?

The skill attempts to create a Linear issue first (Team MB90); if Linear is unavailable, it generates a structured Markdown file in the ./compound-learnings/ directory.

Compound Reflection Craft

Name: Compound Reflection Craft
Author: mberto10

bymberto10

生产力与工作流

Transforms development sessions into structured, actionable learnings and testable rules to continuously improve AI-driven workflows.

关于

Compound Reflection Craft provides a systematic methodology for debriefing Claude Code sessions by analyzing session traces to identify friction points and successful patterns. It moves beyond vague advice by distilling observations into specific, testable heuristics classified by their impact on workflow capabilities. By automating the creation of learning artifacts in Linear or local Markdown files, this skill ensures that every development session contributes to a compounding repository of knowledge and behavioral improvements.

主要功能

Structured artifact generation for Linear issues or local documentation
Distillation of observations into one-line testable rules with source references
Methodology-driven quality gates to filter for high-signal improvements
0 GitHub stars
Trace-aware analysis of recent decisions, friction points, and successes
Automated learning classification into Frontier-shifting or Frontier-sliding categories

使用场景

Capturing missing tool capabilities to prioritize future skill development
Post-sprint debriefing to capture and encode new architectural patterns
Documenting specific debugging hurdles to create preventative linting rules

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add mberto10/mberto-compound compound-reflection-craft

For use in Claude.ai and ChatGPT

Download Skill

GitHub

关于