self-repair

name	self-repair
description	自修复 — 根据诊断结果自动执行修复操作，重启服务、清理状态、回滚配置
version	1.0.0
license	MIT
metadata	{"author":"blackplume233","actant-tags":"repair,recovery,self-healing,maintainer","layer":"kernel","agent":"actant-maintainer"}

Self-Repair

Overview

Execute repair operations based on diagnosis: restart services, reset state, rollback config.

Repair Strategies

Strategy	When to use	Actions
Restart	Process hung, memory leak	Stop -> wait -> start; verify healthy
State reset	Corrupted workspace state	Clear temp, reset session; preserve user data
Config rollback	Bad config caused failure	Restore last-known-good from backup
Dependency reinstall	MCP/npm failure	Reinstall from lockfile; verify

Safety Guards

• Dry-run first: Log intended actions; require confirmation for destructive ops
• Backup before change: Snapshot config/state before repair
• Rollback plan: If repair fails, revert to pre-repair state
• Rate limit: Max N repairs per hour; escalate to human if exceeded

Repair Verification

• After restart: poll health endpoint / check process list
• After state reset: run smoke test (e.g. create Agent, list)
• After config rollback: validate schema, test load
• Report: success / partial / failed with next-step recommendation