| name | self-improvement |
| description | Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Lulu ('No, that's wrong...', 'Actually...'), (3) User requests a capability that doesn't exist, (4) An external API or tool fails, (5) Claude realizes its knowledge is outdated or incorrect. |
Self-Improvement Skill
Log learnings and errors to markdown files for continuous improvement.
When to Log
| Situation | Log To |
|---|
| Command/operation fails | .learnings/ERRORS.md |
| User corrects you | .learnings/LEARNINGS.md |
| User wants missing feature | .learnings/FEATURE_REQUESTS.md |
| API/external tool fails | .learnings/ERRORS.md |
| Knowledge was outdated | .learnings/LEARNINGS.md |
Setup
Create .learnings/ directory in project root:
mkdir -p .learnings
Format
Error Entry
## [ERR-YYYYMMDD-XXX]
**Logged**: 2026-05-01T12:00:00Z
**Status**: pending
**Area**: frontend | backend | infra | tests | docs | config
### Summary
Brief description of what failed
### Error
Actual error message
### Suggested Fix
How to resolve this
Learning Entry
## [LRN-YYYYMMDD-XXX] category
**Logged**: 2026-05-01T12:00:00Z
**Status**: pending
**Area**: ...
### Summary
One-line description
### Details
What happened, what was wrong, what's correct
### Suggested Action
Specific fix or improvement
Feature Request
## [FEAT-YYYYMMDD-XXX]
**Logged**: 2026-05-01T12:00:00Z
**Status**: pending
**Priority**: medium
### Requested Capability
What user wanted
### Complexity Estimate
simple | medium | complex
Status Values
pending - Not yet addressed
in_progress - Being worked on
resolved - Fixed
wont_fix - Decided not to address
promoted - Added to CLAUDE.md, AGENTS.md, or other docs
Promotion
When learning applies broadly, promote to:
CLAUDE.md - Project facts and conventions
AGENTS.md - Workflows and automation rules
Quick Actions
grep -h "Status**: pending" .learnings/*.md | wc -l
grep -B5 "Priority**: high" .learnings/*.md | grep "^## \["
Best Practices
- Log immediately - context is freshest right after issue
- Be specific - include file paths and error messages
- Include reproduction steps - especially for errors
- Suggest concrete fixes - not just "investigate"
- Review
.learnings/ before major tasks