Improve Skill — Development skill for Claude Code
Propose and apply targeted improvements to a specific Claude Code skill (self-improving skills AMEND layer).
How to install Improve Skill
Installs to ~/.claude/skills/unisone-self-improving-skills-improve-skill/SKILL.md
mkdir -p ~/.claude/skills/unisone-self-improving-skills-improve-skill && curl -fsSL https://raw.githubusercontent.com/unisone/self-improving-skills/HEAD/commands/improve-skill.md -o ~/.claude/skills/unisone-self-improving-skills-improve-skill/SKILL.md Restart Claude Code, or start a new session, for it to be picked up.
What Improve Skill does
description: Propose and apply targeted improvements to a specific Claude Code skill (self-improving skills AMEND layer) argument-hint: allowed-tools: Read, Glob, Grep, Write, Edit
Improve Skill: $ARGUMENTS (Self-Improving Skills — AMEND)
You are proposing targeted improvements to the skill `$ARGUMENTS`. This is the AMEND step of the Observe→Inspect→Amend→Evaluate loop. Changes must be evidence-based, minimal, and reversible.
Step 1: Read the Current Skill
Read `~/.
Alternatives in Development
- Grammar Check — Identify grammar and flow errors with targeted fix suggestions 7.8k ★
- AgentHandover — What if OpenClaw, Claude Code, Codex etc. knew how to do your work without you saying it 600 ★
- 20x — Self-improving Agent orchestrator for all knowledge work 107 ★
Full documentation available on GitHub
View Source RepositoryRelated Skills
Skill Rollback
Roll back a skill to a previous version (self-improving skills AMEND layer — undo)
Auto Improve
Cycle d'amelioration automatique - Enchaine retro + apply + compact + cross-scan pour un systeme self-improvin
Meta Improve
Analyze evaluator feedback and agent run failures, then rewrite command templates to prevent recurring issues.
Adw Eval
Measure the loop itself, then PROPOSE (never silently apply) up to 3 improvements
Ilk Self Improve
Plan toolkit improvements from the shared improvement backlog.
Skill Evaluate
Evaluate whether a skill amendment improved outcomes (self-improving skills EVALUATE layer)
Related Agents
Improver
Dispatched by self-improve, self-lint, self-perf, and refactor workflows to apply targeted code improvements (
Config Manager
Owns the global Claude Code setup and its self-improvement loop. Use via the /update-config skill (or directly
Rnd
Use this agent for RESEARCH & DEVELOPMENT — scanning the MCP ecosystem, competitor directories, upstream stack