OpenClaw — agent identity and memory setup

general

openclaw-prompts-and-skills is a Claude Code skill that brings the OpenClaw AI agent's personality and behavior to Claude Code using plain markdown files — no model fine-tuning required. The repository includes seven files: SOUL.md defines the agent's values and opinions, IDENTITY.md sets its name and vibe, USER.md stores user context, CLAUDE.md governs workspace behavior and memory rules, TOOLS.md captures environment details, BOOTSTRAP.md runs a first-conversation ritual, and OPENCLAW_SYSTEM_PROMPT_STUDY.md documents exactly how OpenClaw assembles its system prompt. Memory works without vector databases — the agent reads and writes `memory/YYYY-MM-DD.md` and `MEMORY.md` as it learns. A proof-of-concept Telegram bot (~100 lines of Python) is included to show how to wire a messaging channel to Claude Code in headless mode with session continuity via `--resume`. Ideal for developers who want Claude Code to maintain a consistent personality, long-term memory, and group-chat etiquette across conversations.