# Self Improvement

> Capture corrections, command failures, tool surprises, missing capabilities, and better approaches as durable project learnings. Use this skill when a task reveals reusable knowledge that should prevent future mistakes or improve agent behavior.

- Skill: `michaelschecht/self-improvement` (Agent Skill)
- Install (CLI): `npx skillmds@latest add michaelschecht/self-improvement`
- Raw SKILL.md: https://api.skillmd.com/api/skills/michaelschecht/self-improvement/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: michaelschecht (https://skillmd.com/u/michaelschecht)
- Updated: 2026-09-08
- Page: https://skillmd.com/skills/michaelschecht/self-improvement

---


# Self Improvement

## Overview
Capture corrections, command failures, tool surprises, missing capabilities, and better approaches as durable project learnings. Use this skill when a task reveals reusable knowledge that should prevent future mistakes or improve agent behavior.

## When to Use
- A command, API call, or integration fails in a non-obvious way.
- The user corrects an assumption or gives project-specific guidance.
- A recurring problem suggests a reusable rule, checklist, or automation.
- A workaround should be promoted into `CLAUDE.md`, `AGENTS.md`, or another project memory file.

## Workflow
1. Decide whether the event is a learning, error, or feature request.
2. Create or update `.learnings/LEARNINGS.md`, `.learnings/ERRORS.md`, or `.learnings/FEATURE_REQUESTS.md`.
3. Include a short summary, context, priority, affected area, suggested action, and related files.
4. Search existing entries before adding a new one; link related entries with `See Also`.
5. Promote broadly useful items into project guidance after they are verified or recur.

## Entry Template
```markdown
## [LRN-YYYYMMDD-001] concise-category

**Logged**: ISO-8601 timestamp
**Priority**: low | medium | high | critical
**Status**: pending
**Area**: frontend | backend | infra | tests | docs | config

### Summary
One-line description.

### Details
What happened, what was misunderstood, and what is now known.

### Suggested Action
Specific prevention or improvement.

### Metadata
- Source: conversation | error | user_feedback | investigation
- Related Files: path/to/file.ext
- Tags: tag1, tag2
- See Also: optional related IDs
```

## Promotion Rules
Promote a learning when it applies across multiple tasks, prevents repeated mistakes, documents an important project convention, or has been verified by a fix. Keep promoted guidance short and action-oriented.

## Source Notes
Adapted from the SkillsMP `self-improvement` skill, selected because it is highly starred on SkillsMP and fills a library gap around persistent agent learning workflows.