Self-Repair System — Autonomous AI Automation

Automated self-diagnostics: restarts crashed Ollama, repairs broken configs, verifies workspace integrity, recovers from common failures.

This skill appears to do what it says: local health checks, repair from backups, and restarting a local Ollama instance. Before installing or enabling it: 1) Review and set backupPaths to directories you trust (the skill will copy files from backups into your workspace). 2) Ensure ollamaHost/ollamaPort remain set to localhost (the code defaults to localhost but can be configured to any host — pointing it at an external host would allow network traffic outside your machine). 3) Be aware it may start a detached local process if Ollama is missing; confirm the ollama binary paths it searches are appropriate for your system. 4) Run first in a controlled/sandbox environment and inspect repair-log.json to verify behavior. If you need stronger guarantees about network isolation, run the skill in an environment that enforces localhost-only network rules.

Type: OpenClaw Skill Name: self-repair-system Version: 1.3.0 The 'self-repair-system' bundle provides a framework for automated workspace maintenance, such as restarting Ollama and restoring corrupted configurations. It includes high-risk capabilities in 'src/self-repair.js', specifically 'runCommand' (arbitrary command execution) and 'killProcess' (process termination). Although these functions implement security mitigations—such as blocking shell injection operators and using 'spawn' without a shell—they represent a significant attack surface. While the code appears to be a legitimate utility with no evidence of malicious intent or data exfiltration, the inclusion of these powerful primitives without an immediate internal use case warrants a cautious classification.