How do I install the Skill Improve skill?

Run npx skillmds@latest add yrzhe/skill-improve in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Skill Improve safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Skill Improve?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Skill Improve free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Skill Improve?

Yrzhe (@yrzhe) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Skill Improve by yrzhe — Agent Skill

Name: Skill Improve
Rating: 4.2 (5 reviews)
Author: Yrzhe

Skill Improve

Improve existing skills after use. When Claude uses a skill and discovers issues (outdated steps, missing edge cases, unclear instructions, wrong assumptions), suggest "/skill-improve" after the current task completes. Also use when user says "fix this skill", "this skill is wrong", "update skill", or "/skill-improve <skill-name>".

Yrzhe Updated Sep 22, 2026

File contents

Skill Improve — Post-Use Skill Maintenance

Improve existing skills based on real usage experience. Always finish the current task first.

Trigger Rules

After task completion: If a skill was used and had issues, suggest: "The <skill-name> skill had some gaps — want me to /skill-improve it?"
Manual: /skill-improve <skill-name> or /skill-improve (defaults to most recently used skill in session)
NEVER interrupt the current task to improve a skill. Finish first, improve after.

Improvement Flow

1. Identify Target

Determine which skill to improve:

If <skill-name> provided, use that
If no argument, use the most recently invoked skill in this session
If no skill was used, ask the user which skill to improve

2. Read Current Skill

Read the target skill's SKILL.md and any bundled resources (scripts/, references/, assets/).

3. Diagnose Issues

Compare the skill's instructions against actual execution experience. Categorize problems:

Category	Example
Outdated	Tool/API changed, deprecated method
Missing	Edge case encountered that skill didn't cover
Redundant	Steps Claude already knows, wasting context
Unclear	Ambiguous instruction that caused wrong action
Wrong	Incorrect assumption or factual error

4. Generate Diff Preview

Show the user exactly what will change:

Quote the original text
Show the proposed replacement
Explain why for each change

Do NOT modify the file until user confirms.

5. Apply Changes

After confirmation:

Edit the SKILL.md (and bundled resources if needed)
Preserve the skill's name and description structure unless the issue is there
Keep changes minimal — fix what's broken, don't rewrite what works

6. Log the Improvement

Update or create ~/.claude/skills/learn/learned/<skill-name>.md:

If the file exists, append to the improvement record section:

### <YYYY-MM-DD>
- **Problem**: <what was wrong>
- **Fix**: <what was changed>
- **Source**: <task context that revealed the issue>

If the file doesn't exist (non-learned skill being improved for the first time), create it:

---
skill: <skill-name>
created: <YYYY-MM-DD>
last_updated: <YYYY-MM-DD>
source: improvement
---

## Improvement Record
### <YYYY-MM-DD>
- **Problem**: <what was wrong>
- **Fix**: <what was changed>
- **Source**: <task context that revealed the issue>

7. Memory Sync (Optional)

If a memory system is available and the fix revealed a universal insight (not just a skill-specific correction), save a concise one-line version to memory.

Scope Limits

Small fixes: Apply directly (typos, outdated commands, missing steps)
Major redesign: If the skill's entire approach is wrong, suggest using /learn to extract a new skill instead of patching
Don't bloat: If adding content would push SKILL.md over 500 lines, suggest splitting into reference files instead

---name: skill-improvedescription: Improve existing skills after use. When Claude uses a skill and discovers issues (outdated steps, missing edge cases, unclear instructions, wrong assumptions), suggest "/skill-improve" after the current task completes. Also use when user says "fix this skill", "this skill is wrong", "update skill", or "/skill-improve <skill-name>".---# Skill Improve — Post-Use Skill MaintenanceImprove existing skills based on real usage experience. Always finish the current task first.## Trigger Rules- **After task completion**: If a skill was used and had issues, suggest: "The `<skill-name>` skill had some gaps — want me to `/skill-improve` it?"- **Manual**: `/skill-improve <skill-name>` or `/skill-improve` (defaults to most recently used skill in session)- **NEVER interrupt** the current task to improve a skill. Finish first, improve after.## Improvement Flow### 1. Identify TargetDetermine which skill to improve:- If `<skill-name>` provided, use that- If no argument, use the most recently invoked skill in this session- If no skill was used, ask the user which skill to improve### 2. Read Current SkillRead the target skill's `SKILL.md` and any bundled resources (`scripts/`, `references/`, `assets/`).### 3. Diagnose IssuesCompare the skill's instructions against actual execution experience. Categorize problems:| Category | Example ||----------|---------|| **Outdated** | Tool/API changed, deprecated method || **Missing** | Edge case encountered that skill didn't cover || **Redundant** | Steps Claude already knows, wasting context || **Unclear** | Ambiguous instruction that caused wrong action || **Wrong** | Incorrect assumption or factual error |### 4. Generate Diff PreviewShow the user exactly what will change:- Quote the original text- Show the proposed replacement- Explain why for each changeDo NOT modify the file until user confirms.### 5. Apply ChangesAfter confirmation:- Edit the SKILL.md (and bundled resources if needed)- Preserve the skill's name and description structure unless the issue is there- Keep changes minimal — fix what's broken, don't rewrite what works### 6. Log the ImprovementUpdate or create `~/.claude/skills/learn/learned/<skill-name>.md`:If the file exists, append to the improvement record section:```markdown### <YYYY-MM-DD>- **Problem**: <what was wrong>- **Fix**: <what was changed>- **Source**: <task context that revealed the issue>```If the file doesn't exist (non-learned skill being improved for the first time), create it:```markdown---skill: <skill-name>created: <YYYY-MM-DD>last_updated: <YYYY-MM-DD>source: improvement---## Improvement Record### <YYYY-MM-DD>- **Problem**: <what was wrong>- **Fix**: <what was changed>- **Source**: <task context that revealed the issue>```### 7. Memory Sync (Optional)If a memory system is available and the fix revealed a universal insight (not just a skill-specific correction), save a concise one-line version to memory.## Scope Limits- **Small fixes**: Apply directly (typos, outdated commands, missing steps)- **Major redesign**: If the skill's entire approach is wrong, suggest using `/learn` to extract a new skill instead of patching- **Don't bloat**: If adding content would push SKILL.md over 500 lines, suggest splitting into reference files instead

Frequently asked questions