domains / families / agent-tooling / skill-improver

skill-improver

combined

This skill should be used when the user asks to "apply skill improvements", "update skill from plan", "execute improvement plan", "fix skill issues", "implement skill recommendations", or mentions applying improvements from quality review reports. Reads improvement-plan-{name}.md files generated by skill-quality-reviewer and intelligently merges and executes the suggested changes to improve Claude Skills quality. Iteratively improve a Claude Code skill using the skill-reviewer agent until it meets quality standards. Use when improving a skill with multiple quality issues, iterating on a new skill until it meets standards, or automated fix-review cycles instead of manual editing.

When to use

• "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"
• 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

When not to use

• **One-time review**: Use `/skill-reviewer` directly instead
• **Quick single fixes**: Edit the file directly
• **Non-skill files**: Only works on SKILL.md files
• **Experimental skills**: Manual iteration gives more control during exploration
• Use this skill only when the task clearly matches the scope described above.
• Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
• Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.
• **One-time review**: dispatch the `plugin-dev:skill-reviewer` agent directly

Skill page

Conflict

The three copies describe three different procedures for improving a skill. The first (trailofbits) is a wrapper that resolves a path and launches a background workflow (improve.js) with a reviewer agent, ledger, and scope guard. The second (sickn33) is a manual loop where the agent directly calls the skill-reviewer, categorizes issues, fixes them, and uses a specific completion marker. The third (brycewang-stanford) is a plan-executor that reads a pre-generated improvement-plan file, parses it, resolves conflicts, and applies changes with backups. Because they rely on different inputs (workflow vs. direct agent call vs. plan file) and have distinct mechanisms (background loop vs. interactive loop vs. batch execution), crowning one would hide the other valid approaches.