skill-improver
**skill-improver** reads improvement-plan-{name}.md files generated by skill-quality-reviewer and automatically executes the suggested changes to Claude Skills by parsing improvement items, grouping changes by file, detecting conflicts, and applying updates through backup and verification. Use this skill when users request applying improvements from a quality review report, updating skills based on identified issues, or executing structured fix recommendations from the skill-quality-reviewer workflow.
git clone --depth 1 https://github.com/Galaxy-Dawn/claude-scholar /tmp/skill-improver && cp -r /tmp/skill-improver/skills/skill-improver ~/.claude/skills/skill-improverSKILL.md
# Skill Improver
Execute improvement plans generated by `skill-quality-reviewer` to automatically update and fix issues in Claude Skills.
## Core Workflow
```
Read improvement-plan-{name}.md
↓
Parse improvement items (High/Medium/Low priority)
↓
Group changes by file
↓
Detect and resolve conflicts
↓
Backup original files
↓
Execute updates (Edit or Write tools)
↓
Verify results
↓
Generate update-report
```
## When to Use
**Trigger phrases:**
- "Apply improvements from improvement-plan-git-workflow.md"
- "Update my skill based on the quality report"
- "Execute the improvement plan for api-helper"
- "Fix the issues identified in quality review"
**Use this skill when:**
- Applying improvements from an improvement plan
- Updating a skill based on quality review feedback
- Executing recommended fixes from `skill-quality-reviewer`
- Implementing structured improvements to skill documentation
## Step-by-Step Guide
### Step 1: Load the Improvement Plan
Read the `improvement-plan-{skill-name}.md` file generated by `skill-quality-reviewer`.
```bash
# Plan is typically in current directory
ls improvement-plan-*.md
# Or specify full path
read /path/to/improvement-plan-my-skill.md
```
**Validate the plan:**
- File exists and is readable
- Contains priority sections (High/Medium/Low)
- Has structured improvement items
- Includes file paths and suggested changes
See `references/plan-format.md` for detailed plan structure.
### Step 2: Parse and Group Changes
Extract all improvement suggestions and organize by target file.
**Extract from each item:**
- File path (e.g., `SKILL.md:line:line` or `references/file.md`)
- Dimension (Description Quality, Content Organization, etc.)
- Impact (+X points)
- Current content
- Suggested content
- Reason for change
**Build update queue:**
```
by_file = {
"SKILL.md": [change1, change2, ...],
"references/guide.md": [change3, ...],
"examples/demo.md": [change4, ...],
}
```
### Step 3: Detect and Resolve Conflicts
Check for conflicts when multiple changes affect the same content.
**Resolution strategy:**
1. High priority takes precedence over medium/low
2. If same priority, preserve first change
3. Flag conflicts for manual review
4. Document resolution in update report
See `references/merge-strategies.md` for detailed merge logic.
### Step 4: Sort by Priority
Order changes by priority within each file.
**Priority order:**
1. High Priority (execute first)
2. Medium Priority (execute second)
3. Low Priority (execute last)
### Step 5: Backup and Execute
**Backup location:** `~/.claude/skills/backup/{skill-name}-{timestamp}/`
```bash
# Use backup script
~/.claude/skills/skill-improver/scripts/backup-skill.sh <skill-path>
```
**Apply changes:**
- Use Edit tool for existing content
- Use Write tool for new files
- Verify each change was applied
### Step 6: Verify and Report
**Verification checks:**
1. YAML syntax valid
2. All modified files exist and are valid
3. New files were created successfully
4. No unintended changes occurred
```bash
# Use verify script
~/.claude/skills/skill-improver/scripts/verify-update.sh <skill-path>
```
**Generate update-report-{skill-name}-{timestamp}.md** documenting:
- Summary (files modified, files created, total changes)
- Changes Applied (per-file breakdown)
- Quality Improvement (before/after scores)
- Verification Results
- Backup Location
See `examples/update-report-example.md` for report template.
## Priority Handling
### High Priority
Execute first. These typically address:
- Critical description issues
- Major writing style problems
- Missing structural elements
- Security-related concerns
### Medium Priority
Execute after High. These typically address:
- Content organization improvements
- Additional examples
- Documentation enhancements
### Low Priority
Execute last. These typically address:
- Minor clarifications
- Nice-to-have improvements
- Polish and refinement
## Integration with Skill Quality Reviewer
This skill works seamlessly with `skill-quality-reviewer`:
```
Current Skill (67/100 D+)
↓ [skill-quality-reviewer]
Improvement Plan
↓ [skill-improver]
Improved Skill (87/100 B+)
↓ [skill-quality-reviewer]
Quality Report (validation)
```
**Iterate until desired quality level reached.**
## Additional Resources
### Reference Files
- **`references/plan-format.md`** - Improvement plan file structure and format
- **`references/merge-strategies.md`** - Detailed merge algorithms and conflict resolution
- **`references/error-handling.md`** - Error handling strategies
- **`references/supported-updates.md`** - Supported update types with examples
### Example Files
- **`examples/improvement-plan-example.md`** - Sample improvement plan
- **`examples/update-report-example.md`** - Sample update report
### Scripts
- **`scripts/backup-skill.sh`** - Create backup of skill before updates
- **`scripts/verify-update.sh`** - Verify skill integrity after updates
## Best Practices
### Before Applying Updates
1. **Review the improvement plan** - Understand what will change
2. **Verify backup location** - Ensure backups can be restored
3. **Check for manual changes** - Note any uncommitted local modifications
4. **Estimate impact** - Review expected quality improvement
### During Update Execution
1. **Process in priority order** - High → Medium → Low
2. **Verify after each file** - Check updates were applied correctly
3. **Log all changes** - Document what was modified
4. **Handle conflicts gracefully** - Flag for review if needed
### After Applying Updates
1. **Review the update report** - Confirm all changes were intended
2. **Test the skill** - Verify it still works correctly
3. **Compare scores** - Check quality improvement matches expectations
4. **Keep backup** - Retain backup until confident in changes
## Usage Examples
**Example 1: Apply improvements to local skill**
```
User: "Apply improvements from improvement-plan-git-workflow.md"
[Claude exExpert code review specialist. Proactively reviews code for quality, security, and maintainability. Use immediately after writing or modifying code. MUST BE USED for all code changes.
Use this agent when the user provides a Kaggle competition URL or asks to learn from Kaggle winning solutions. Examples:
Use this agent when the user asks to "conduct literature review", "search for papers", "analyze research papers", "identify research gaps", "review related work", or mentions starting a research project. This agent integrates with Zotero for automated paper collection, organization, and full-text analysis. Examples:
Use this agent when the user provides a research paper (PDF/DOCX/arXiv link) or asks to learn writing patterns from papers, extract venue-specific writing signals, study paper structure, or mine rebuttal strategies. The agent writes extracted knowledge into the active installed paper-miner writing memory for ml-paper-writing. It does not maintain project-specific writing memory.
Use this agent when the user asks to "write rebuttal", "respond to reviewers", "analyze review comments", or needs help with academic paper review response. This agent specializes in systematic rebuttal writing with professional tone and structured responses.
Test-driven development guide for writing tests first, implementing the smallest passing change, and keeping verification tight. Use when the user explicitly wants TDD or when a task should be driven by failing tests before code.
Run a blocker-first post-experiment workflow: validate evidence, produce strict statistical analysis when possible, and generate a decision-oriented results report only when the analysis bundle is sufficient. Uses results-analysis + results-report as a gated two-stage workflow.
Commit changes following Conventional Commits format (local only, no push).