xiaohua-self-improving

What This Skill Does

xiaohua-self-improving is a specialized agent framework tailored for the OpenClaw environment, specifically optimized for users operating within domestic (Chinese) infrastructure. It functions as a meta-cognitive layer for your agent, enabling it to learn from past errors, store long-term knowledge in MEMORY.md, and follow a structured 'Look-Think-Find-Do' workflow. By formalizing error tracking and knowledge consolidation, it transforms your agent from a stateless assistant into an evolving entity that retains institutional knowledge across sessions.

Installation

To integrate this skill into your environment, use the ClawHub package manager:

npx clawhub install xiaohua-self-improving

After installation, ensure your workspace has the necessary directory structure, specifically the .learnings/ folder for logs and MEMORY.md for core knowledge synthesis.

Use Cases

• Continuous Integration of Wisdom: Automatically distill technical trial-and-error sessions into durable documentation.
• Error Management: Standardize the logging of system faults using the ERR-XXX convention to prevent repetitive failures.
• Agent Behavior Tuning: Use the SOUL.md linkage to refine the agent's personality and decision-making logic over time.
• Domestic Production Efficiency: Seamlessly connect to Baidu Qianfan APIs and local RAG models without reliance on restricted international infrastructure.

Example Prompts

1. "Look at the recent errors in .learnings/ERRORS.md and propose a plan to update our MEMORY.md with a permanent fix."
2. "Perform the 'Look-Think-Find-Do' cycle for the task of migrating our local database, starting by reviewing the existing context in our memory files."
3. "Summarize the last three days of activity from memory/ folder and suggest a new rule for SOUL.md to improve my output quality."

Tips & Limitations

• Append, Never Overwrite: Always use append-only protocols for daily memory files to maintain an audit trail.
• Consistency: Use the standardized ERR-XXX formatting for all error logs to ensure the agent can regex-parse them effectively.
• Resource Usage: Frequent file I/O operations in the workspace can impact performance on low-end hardware; monitor your disk activity if you have an extensive memory history.
• Dependency: This skill works best when paired with local-file-rag-basic for optimal search latency.