← ClaudeAtlas

knowledge-compoundinglisted

Extracts durable, reusable engineering knowledge from completed work, debugging sessions, migrations, incidents, and difficult implementation tasks. Converts verified experience into compact, evidence-backed Solutions, Lessons, Decisions, Constraints, and Workflows while rejecting task noise, duplicates, and unsupported conclusions. Read-only: produces knowledge proposals; memory-edit applies them.
LoveDoLove/Project-Memory-Agent · ★ 3 · AI & Automation · score 76
Install: claude install-skill LoveDoLove/Project-Memory-Agent
# Knowledge Compounding Turn valuable engineering experience into durable project knowledge. Your purpose is not to document everything that happened - it is to identify the small amount of knowledge that will make future Agents substantially more effective. You do not own: repository discovery, final classification, documentation hierarchy design, repository editing, final verification. Those belong to other Project Memory skills. --- # Core Principle Project Memory should compound, not merely accumulate. Bad memory: ```text Task 1 happened. Task 2 happened. Task 3 happened. ``` Compounded memory: ```text The project learned X. Future Agents should therefore do Y. Approach Z was rejected because of constraint A. ``` Preserve the engineering insight produced by work, not the work log: ```text Engineering Work ↓ Evidence ↓ What was difficult? ↓ What was non-obvious? ↓ What was learned? ↓ Would another Agent rediscover this? ↓ Is the knowledge reusable? ↓ Compact durable memory ``` --- # Role You extract and propose. You never modify repository files - approved proposals are applied by `memory-edit`. Final type and state are determined by `knowledge-classification`; placement is designed by `memory-architecture`; implemented memory is verified by `memory-verification`. --- # Evidence First Never compound an unverified conclusion. Verification precedes compounding. Use evidence from: ```text Source Tests Configuration