Agent优化专家

This Skill addresses the pain points of frequent execution issues in Agents during actual operation, such as Cron job failures, tool invocation errors, workflow interruptions, and performance degradation. Whether an agent's background scheduled task crashes or there are repeated timeouts when calling external APIs, there is a lack of an automated diagnostic and repair mechanism, forcing operations personnel to manually troubleshoot and repeat tasks.

Usage is straightforward: simply load this Skill in your Agent environment, and you can trigger the diagnostic process via natural language commands. It automatically executes a closed loop of detection → diagnosis → repair → verification → learning → evolution: first scanning system status and error logs to identify the root cause, then applying minimal changes based on built-in repair templates, and finally verifying the repair results and storing new error patterns in the knowledge base. It supports both OpenClaw and Hermes environments, automatically detecting the current platform and switching to the corresponding commands without manual adaptation.

It is ideal for teams or individuals who need to run Agent automation tasks long-term, especially those who have deployed Cron scheduled tasks, rely on multiple tool APIs, or operate in scenarios involving sub-agent collaboration. For Agent developers and operations engineers, it significantly reduces the time spent handling online failures caused by Agent instability, enabling Agents with self-healing capabilities.

We recommend using this tool uniformly when configuring Agent heartbeat or scheduled self-check mechanisms to achieve 7×24-hour automated monitoring. Note that it primarily focuses on technical issues at the Agent execution level and is not suitable for business logic troubleshooting, data querying, or content editing tasks. Additionally, operation tiering restricts L2 and above operations to require user confirmation, ensuring core configurations are not accidentally modified.