Skip to main content
skill CX Codex

self-improvement — Claude Code skill: log fixes and errors

Keeps your corrections, failed commands and feature asks in markdown files so the agent can turn repeated lessons into project rules.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add pskoett/pskoett-ai-skills --skill self-improvement -a claude-code
Open on GitHub

What self-improvement does

The skill sets up a .learnings/ folder at the project root and sorts entries into three files: LEARNINGS.md for corrections, knowledge gaps and best practices, ERRORS.md for commands and external tools that failed, FEATURE_REQUESTS.md for capabilities the project lacks. Every entry gets an ID like LRN-20250115-001, plus priority, status and a metadata block. Once an issue is closed, the status flips to resolved and a Resolution section records the commit and notes.

File templates live in assets/ and get copied into .learnings/, or you write the headers yourself. Each entry also carries context-maintenance fields at capture time: how fast the note decays, where the claim came from, how to recheck it. Open questions stay marked, they just do not get dropped.

This is the passive layer. Fixing a live failure mid-task belongs to the companion skill self-healing, which files verified HEAL- entries to .learnings/HEALS.md. self-improvement collects the repeats and promotes a stable rule into CLAUDE.md, AGENTS.md or .github/copilot-instructions.md. Both share .learnings/, each writes its own files. No folder in the root means nowhere to write, so create it first.

What it helps with

  • record a user correction so the agent skips that mistake next session
  • capture a failed command with its output and environment details
  • file a feature request with a rough complexity estimate
  • flag outdated agent knowledge and track its status until fixed
  • promote a recurring pattern from .learnings/ into project rules

Example prompts for the agent

  • Log my fix about date formatting to LEARNINGS.md under correction
  • Add the failed npm run build to .learnings/ERRORS.md — missing env var
  • Review .learnings/ before we refactor the billing module

How to install self-improvement

1Skills CLI (Vercel)

npx skills add pskoett/pskoett-ai-skills --skill self-improvement -a claude-code

2Manually: put the files in place

~/.claude/skills/self-improvement/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/self-improvement/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about self-improvement

How do I install self-improvement in Claude Code?

Run npx skills add pskoett/pskoett-ai-skills --skill self-improvement -a claude-code. The skill lands in ~/.claude/skills/self-improvement/SKILL.md.

What does self-improvement do?

It keeps a journal of corrections, errors and feature requests under .learnings/ and helps move repeated findings into the project's rule files.

How is it different from self-healing?

self-healing repairs an active failure during a task and files verified HEAL- entries to .learnings/HEALS.md. self-improvement gathers those repeats and, once a pattern holds, turns it into a rule or a new skill.

Similar skills

All skills for Agent setup: memory, context, orchestration →

From the same set4

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor