메뉴
Update an existing skill's SKILL.md content or frontmatter. Use when the user says "update skill", "change skill", "skill güncelle", "modify skill", or "update-skill <name>".
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | update-skill |
| description | Update an existing skill's SKILL.md content or frontmatter. Use when the user says "update skill", "change skill", "skill güncelle", "modify skill", or "update-skill <name>". |
| argument-hint | <skill-name> |
| allowed-tools | Read, Glob, Write, Edit |
| disable-model-invocation | true |
Target skill: $ARGUMENTS
If no argument given, Glob .claude/skills/ for directories containing SKILL.md, then list them and ask which one to update.
Otherwise, look for .claude/skills/<skill-name>/SKILL.md. If not found, stop and say so.
Read the SKILL.md file. Print the current frontmatter and workflow to the user so they can see what they're changing.
Current skill: <skill-name>
What would you like to change?
1. Frontmatter (name, description, triggers, allowed-tools)
2. Workflow steps
3. Both
Also: Why is this changing?
Wait for answer.
Edit .claude/skills/<skill-name>/SKILL.md with the requested changes.
Preserve conventions:
disable-model-invocation: true unless user explicitly asks to remove itallowed-tools to minimum required set$ARGUMENTS placeholder where the skill accepts inputAfter editing, verify:
argument-hint is present if skill uses $ARGUMENTSallowed-toolsIf anything is inconsistent, fix it and inform the user.
Updated: .claude/skills/<skill-name>/SKILL.md
Changed: <frontmatter / workflow / both>
Reason: <user's reason>
Analyze accumulated feedback memories and turn recurring patterns into permanent controls (rules, permissions, hooks) or common-mistakes entries. Use when the user says "improve", "iyileştir", "apply feedback", "kurallaştır", or "turn feedback into rules".
Project onboarding and template setup. Use when the user says onboard, setup, initialize, init, configure this project, or first-time setup. Guides through detecting project mode and stack, configuring CLAUDE.md, permissions, hooks, rules, and architecture docs.
Save session learnings, patterns, and corrections to persistent memory. Use when the user says "save learnings", "save memory", "remember this session", or at the end of a productive session.
Add a new Claude Code control following project conventions. Use when the user says "new rule", "add rule", "create rule", "yeni rule ekle", or wants to enforce behavior consistently. Guides through choosing enforcement surfaces, writing effective rule files, and keeping controls consistent.
Create a new Claude Code skill following project conventions. Use when the user says "new skill", "add skill", "create skill", or "yeni skill ekle". Guides through gathering requirements, writing effective SKILL.md files, and structuring skill directories.
Add a new Claude Code hook — shell script wired into settings.json. Use when the user says "new hook", "add hook", "yeni hook ekle", or wants to automate something that runs on every tool use, session start, or stop.