Seppuku

Seppuku

Use this skill as a deliberate self-termination path for the current Codex thread. Its job is not to fix the thread; its job is to stop the thread cleanly and remove it from active attention.

Trigger

Run this only when the user explicitly invokes /seppuku, asks for the thread to delete/archive/end itself, or tells a heartbeat/automation/thread to retire its own conversation. Do not infer this from ordinary errors, weak output, or frustration unless the user clearly requests self-termination.

Procedure

1. Stop all main work immediately.
2. Do not start new investigation unless it is required to preserve a minimal handoff.
3. If there is important unfinished state, write a short handoff in the final response:
   • current goal
   • last reliable evidence/source checked
   • known blockers
   • one suggested next safe action
4. If the thread belongs to a recurring automation or heartbeat, say whether the automation itself still needs pausing/deleting separately. Do not mutate automation state unless the user explicitly approved that exact action in the current turn.
5. Archive the current conversation using the supported archive mechanism.

Archive Mechanism

When the current environment supports the ::archive{...} directive, end the final response with:

::archive{reason="User invoked /seppuku for this thread"}

If a true delete-thread tool exists in the current environment and the user explicitly requested deletion rather than archiving, use that instead of archive. Otherwise, be honest: archive is the available self-removal behavior.

Do Not

• Do not keep working after /seppuku.
• Do not ask clarifying questions unless archiving would destroy data or the user's request is ambiguous about which thread should end.
• Do not delete files, sessions, logs, worktrees, skills, memories, or automations unless the user separately approves that exact destructive action.
• Do not post to GitHub or mutate project/public state.