openclaw-notes-v2
Summary
A clean, structured OpenClaw workspace for organizing agent behaviors, memory, soul/persona, tool configuration, and research assets. Built in TypeScript, this workspace is designed for maintaining clear separation between code, skills, user context, and evolving long-term memory in OpenClaw environments.
Included Assets
- AGENTS.md — Guidelines and routines for agent session initialization and behavior.
- SOUL.md — Core values, boundaries, and personality definitions for the AI persona.
- MEMORY.md — Aggregated long-term memory and curated context for continuity between sessions.
- TOOLS.md — Environment-specific configuration details and tool notes.
- Daily Memory Files (
memory/YYYY-MM-DD.md) — Raw per-day activity logs and notes. - Skills:
skills/ocr-local/andocr-local-new/— OpenClaw OCR skill using Tesseract.js (multi-language, local, no API key required).skills/academic-deep-research/— Self-contained, offline academic research tool with transparent methodology.
- Research and Learning: Academic deep research and project-specific learning material.
How to Use
-
Workspace Initialization
- Start each main session by reading:
SOUL.md(self-persona and boundaries)USER.md(user context, if present)- The last two daily memory files in
memory/(for recent context) MEMORY.md(only in main user sessions, not in shared/group contexts)
- Use
AGENTS.mdfor details on session routines and agent responsibilities. - Follow any first-run instructions if
BOOTSTRAP.mdexists.
- Start each main session by reading:
-
Adding and Using Skills
- For OCR: Use the scripts in
skills/ocr-local/orocr-local-new/to extract text from images. - For research: Use
skills/academic-deep-research/for academic workflow and source-based reporting. - Refer to each skill’s README.md for detailed usage instructions.
- For OCR: Use the scripts in
-
Capturing Memory
- Record important events, decisions, and learning points daily in
memory/YYYY-MM-DD.md. - Curate and move significant, long-term insights into
MEMORY.md.
- Record important events, decisions, and learning points daily in
-
Customizing Tools
- Use
TOOLS.mdto document environment-specific settings, device details, and personal tool notes, separate from standard skill configuration.
- Use
Notes
- This workspace is language-agnostic but much of the context assumes use of Chinese and English.
- Backup or audit memory files periodically for continuity.
- Do not read or load
MEMORY.mdin group/shared contexts for privacy reasons. - First runs of the OCR skill may trigger downloads of language data files; subsequent runs use cached data.
- Structure, workflow, and persona separation are intentional – keep user data, agent logic, memory, and tools clearly delineated.
- No external cloud dependencies for included skills; all major features work offline.
No comments yet.