szotasz/marveen/skill-management
Claude Code and agent skill by szotasz/marveen
- community
Checking that the SKILL.md file is still in the repository.
List, inspect, patch, or delete skills from ~/.claude/skills/. Use when the user asks about available skills, wants to modify an existing skill, or when a retrospective proposes skill changes. Trigger on "/skills" command or skill-related retrospective actions.
Install
npx vskill@latest install szotasz/marveen/skill-managementSKILL.md
Skill Management -- CRUD for the Skill Library
When to use
- User asks "what skills do I have?" or "list skills"
- User wants to inspect a specific skill's content
- A
/retrospectiveproposes CREATE, PATCH, or DELETE actions on skills - User says "update skill X" or "fix skill Y"
- Periodic audit: check for stale or duplicate skills
Arguments
| Argument | Required | Description | |
|---|---|---|---|
action | YES | One of: list, show, create, patch, delete, audit | |
name | for show/patch/delete | Skill directory name (e.g. github-pr-rebase-merge) | |
scope | no | global (default, ~/.claude/skills/) or local (./claude/skills/) |
Examples:
/skills action=list-- list all skills with descriptions/skills action=show name=github-pr-rebase-merge-- show full SKILL.md/skills action=create name=new-skill-- interactive skill creation/skills action=patch name=existing-skill-- modify specific sections/skills action=delete name=old-skill-- remove with confirmation/skills action=audit-- find stale, duplicate, or oversized skills
Procedure
action=list
Present as a compact table. If both global and local exist, show both.
action=show
- Read
~/.claude/skills/{name}/SKILL.md - If it has a
references/subdirectory, list those files too - Present the full content
action=create
Interactive skill creation workflow:
- Ask for trigger context: "When should this skill activate?"
- Ask for procedure steps: "What does it do, step by step?"
- Generate SKILL.md with proper frontmatter:
- Write to
~/.claude/skills/{name}/SKILL.md - Update
.skill-index.mdif it exists
Rules:
- Keep SKILL.md under 500 lines
- Large reference material goes in
references/subdirectory - Description must be specific enough for L0 matching (not "does stuff")
- Procedure steps must be concrete and executable
action=patch
Targeted modification of an existing skill:
- Read the current SKILL.md
- Identify the section to change based on user input or retrospective proposal
- Apply targeted edit (old text -> new text), not full rewrite
- If adding a pitfall from a runtime discovery, append to the Pitfalls section
- Log the patch reason in the Pitfalls section if it came from an error recovery
Rules:
- Never rewrite the entire skill for a small change
- Preserve existing pitfalls and procedure steps unless explicitly removing
- If the patch changes triggers, update the description in frontmatter too
action=delete
- Show the skill content first
- Ask for confirmation: "Delete skill '{name}'? This removes the entire directory."
- On confirmation:
- Update
.skill-index.mdif it exists
Rules:
- Never delete without showing content first
- Never delete without explicit user confirmation
- If the skill is referenced by other skills, warn before deleting
action=audit
Comprehensive skill library health check:
- Stale detection: Skills not referenced in any CLAUDE.md and with no git activity in 60+ days
- Duplicate detection: Skills with overlapping descriptions or triggers
Review for semantic overlaps manually.
- Size check: Skills over 500 lines that should be refactored
- Index sync: Check
.skill-index.mdmatches actual directories
Present findings as:
Skill Index Regeneration
When skills are created, patched, or deleted, regenerate the index:
Pitfalls
- Do NOT auto-delete skills without user confirmation
- Do NOT create skills for one-off tasks (check the 2+ occurrence rule)
- Audit results are advisory, not auto-executed
- The
.skill-index.mdis for L0 matching only; the full SKILL.md is L1 - On macOS,
statsyntax differs from Linux; the audit commands handle both - If a skill references external APIs or tokens, never include the actual values
Relation to other mechanisms
| Mechanism | Skill-management's role | |
|---|---|---|
| retrospective | Retrospective proposes changes; skill-management executes them | |
| CLAUDE.md | CLAUDE.md references skills; skill-management maintains the library | |
| .skill-index.md | Skill-management regenerates this after mutations | |
| seed-skills/ | Seed skills are templates; once installed they become regular skills managed here |
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingseed-skills/skill-management/SKILL.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/szotasz/marveen/skill-management)