Self-Improving 中文适配

Self-reflection + Self-criticism + Self-learning + Self-organizing memory. Agent evaluates its own work, catches mistakes, and improves permanently. Use when...

This skill is internally coherent for local, persistent self-improvement: it will create and maintain ~/self-improving/ and add small steering snippets to workspace files. Before installing or running it, (1) back up any AGENTS.md / SOUL.md / HEARTBEAT.md you care about so edits can be reviewed; (2) inspect the created files after setup to confirm content and that no secrets were recorded; (3) decline or review carefully before consenting to install the optional 'Proactivity' companion (it performs network installs); (4) if you are privacy-sensitive, consider running in Passive or Strict mode and use the provided 'Never Store' boundaries strictly; and (5) if you later want to remove learned state, use the documented 'forget' / 'forget everything' flow and verify files are removed. Overall the skill looks coherent and expected for its stated purpose.

Type: OpenClaw Skill Name: self-improving-cn Version: 1.2.16 The skill bundle is a comprehensive framework for enabling an AI agent to learn from user corrections and manage its own memory. It uses local file storage in `~/self-improving/` and integrates with the OpenClaw environment via standard configuration files like `SOUL.md` and `HEARTBEAT.md`. A key safety feature is the `boundaries.md` file, which explicitly instructs the agent never to store sensitive information such as credentials, API keys, or PII. The logic is transparent, providing the user with commands to audit, export, or wipe the memory, and it lacks any indicators of data exfiltration, malicious execution, or unauthorized persistence.