mupengism

Enables AI agents to maintain session continuity, develop self-identity, and manage file-based long-term memory anchored by human partners.

This skill is essentially a set of templates and practices for file-based agent memory — before using it: (1) review and customize SOUL.md/SOUL-TEMPLATE.md yourself so nothing sensitive is written; (2) never put secrets (API keys, passwords, mnemonics) into MEMORY/SELF/SOUL files; (3) be careful when editing agent config (AGENTS.md) to auto-load SOUL.md — only do this in trusted agent instances and consider file permissions or encryption if the workspace is synced or shared; (4) remember there is no downloaded code here, so risk is mostly about information exposure and agent behavior driven by those local files.

Type: OpenClaw Skill Name: mupengism Version: 1.0.0 This skill bundle is designed to provide an AI agent with a file-based memory system and a framework for self-identity and continuity. It instructs the agent to read and write to local files within its workspace (e.g., `SOUL.md`, `MEMORY.md`, `memory/L1-active.md`). Crucially, the `PRINCIPLES.md` and `SOUL-TEMPLATE.md` files contain explicit and robust instructions to the AI agent to prevent prompt injection, data exfiltration (e.g., '프라이빗 키/니모닉/토큰 절대 노출 금지'), and unauthorized external actions (e.g., '무단 외부 행동 — 이메일, 트윗 보내기 전 확인'). There is no evidence of malicious intent, data exfiltration to external endpoints, backdoors, or execution of suspicious commands. The file operations are aligned with the stated purpose of local memory management.