메뉴
Archive an agent instead of deleting it — fully reversible. Use when the user says "retire agent", "archive agent", "deactivate agent", "disable agent", or "agenti kaldır".
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 | retire-agent |
| description | Archive an agent instead of deleting it — fully reversible. Use when the user says "retire agent", "archive agent", "deactivate agent", "disable agent", or "agenti kaldır". |
| argument-hint | <agent-name> |
| allowed-tools | Read, Glob, Edit, Bash(git mv *), Bash(mkdir *) |
| disable-model-invocation | true |
Target agent: $ARGUMENTS
If no argument given, Glob .claude/agents/ and list available agents, then ask which to retire.
Otherwise check:
.claude/agents/<name>.md → Type A.claude/agents/<name>/ → Type BIf neither found, stop and say so.
Retiring: <agent-name> (<Type A / Type B>)
This will move the agent to .claude/agents/archive/<agent-name>.
It will no longer be active, but nothing is deleted — you can restore it with /restore-agent <name>.
Proceed? (yes / no)
Wait for confirmation.
Ensure .claude/agents/archive/ exists (create if needed).
Type A:
git mv .claude/agents/<name>.md .claude/agents/archive/<name>.md
Type B:
git mv .claude/agents/<name> .claude/agents/archive/<name>
If git mv is unavailable (untracked files), use Bash mv instead.
Before archiving, check if this agent's AGENT.md Skills table lists any skills:
.claude/agents/<name>/AGENT.md and scan the Skills tableWarning: <agent-name> has skills listed in its AGENT.md Skills table:
- <skill-1>
- <skill-2>
These skills will become orphaned (no parent agent).
Options:
a) Retire those skills too with /retire-skill (recommended)
b) Keep the skills and archive the agent anyway
c) Cancel
Wait for answer. If (a), run /retire-skill for each listed skill.
Retired: <agent-name>
Archived to: .claude/agents/archive/<name>[.md]
[Skills also retired: <list>] ← if applicable
To restore: move back to .claude/agents/
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.