skills.rest
jongensutrecht
jongensutrechtCOMMUNITY
Explore all skills by this author →
GitHub

compound-learning

Capture proven learnings from completed runs into LEARNINGS.md files.

1|Updated Jan 13, 2026
Search Tags:#memory#archiving#deduplication#knowledge capture#audits#learned#run review
One-click install
npx skills add https://github.com/jongensutrecht/demo-ai-stack --skill compound-learning-jongensutrecht
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: compound-learning
Source: https://github.com/jongensutrecht/demo-ai-stack/tree/main/lars%20skills/compound-learning
Command: npx skills add https://github.com/jongensutrecht/demo-ai-stack --skill compound-learning-jongensutrecht

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill captures concrete, proven learnings from completed runs, reviews, and audits so future sessions start smarter and avoid repeating the same mistakes.

Core Features & Use Cases

  • Evidence-based extraction: Filters raw observations down to only actionable learnings backed by real runs.
  • Deduplication and archiving: Updates existing entries instead of duplicating them, and moves overflow into an archive.
  • Cross-session memory: Persists repo-specific and global learnings so later planning, reviews, and audits can reuse them.
  • Use case: After a CTO review or audit, it records the most important patterns, false positives, gaps, and guard checks for future work.

Quick Start

Ask this skill to review the latest run output and append the most valuable deduplicated learnings to the appropriate LEARNINGS.md file.

Frequently Asked Questions about compound-learning

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I capture proven learnings from completed code audits?▼

Yes, you can persist post-run knowledge across sessions by writing captured learnings to repo-specific and global memory files. This allows later planning, reviews, and audits to access and reuse the accumulated knowledge.

What is the best way to deduplicate audit learnings in memory files?▼

You start post-run knowledge management by asking the skill to review the latest run output. It then filters raw observations down to actionable learnings and appends the most valuable deduplicated entries to the appropriate LEARNINGS.md file.

Do I need completed review runs to extract actionable learnings?▼

No, the skill does not support manual knowledge entry. It strictly requires evidence-based extraction from completed runs, reviews, and audits to ensure that only proven patterns, false positives, and guard checks are recorded.

When should I archive overflow entries in learning memory files?▼

You should archive overflow entries in learning memory files when updating existing entries exceeds the defined max-entry limits. This overflow archiving process maintains strict adherence to guard rules while preserving historical knowledge.

Related Skills

Explore More Software Engineering Skills→
JubaKitiashviliJubaKitiashvili
41

retrospective

Evaluate completed coding sessions and generate structured retrospective reports.

Community
Advanced
evanfang0054evanfang0054

session-learnings

Log and retrieve project-specific learnings across sessions with custom scripts.

Community
Intermediate
JustaName-idJustaName-id

jaw-sdk-best-practices

Configure JAW SDK packages for passkeys, permissions, gas sponsoring, and ENS.

Official
Advanced
wormholecowboywormholecowboy
1

stateUpdate

Append conversation progress, learnings, and decisions to timestamped .state files.

Community
Intermediate