Trading AI Workspace
Trading AI assistant — OpenClaw configuration with personality, skills, and stock screener tooling.
Summary
An OpenClaw workspace configured as a trading-focused AI assistant with:
- Personality: Protective caretaker with strong memory focus
- Skills: Kimi IM integration, time-awareness, worker safety constraints
- Memory: Session consolidation system (STM/LTM), diary generation, daily logs
- Tooling: Stock screener generator with hourly GitHub Actions updates
Included Assets
| Path | Purpose |
|---|---|
SOUL.md | Personality, behavior modes (work/casual), diary & easter eggs |
IDENTITY.md | Name, creature type, vibe definition |
USER.md | User profile template (name, timezone, preferences) |
AGENTS.md | Session startup workflow and memory loading rules |
TOOLS.md | Environment-specific notes template |
skills/kimiim/ | Kimi Group Chat integration (listing, replying, threading) |
skills/time-awareness/ | Mandatory session_status call for relative time queries |
skills/worker-safety/ | Hard limits (no runtime upgrades, 127.0.0.1 gateway, no prompt injection) |
memory_consolidation/ | Python scripts for session stats, STM/LTM compact, diary generation |
scripts/generate_screener.py | Stock screener generator using Yahoo Finance data |
stock_screener_header.html | Screener HTML template/styles |
.github/workflows/update-screener.yml | Hourly screener auto-update (Mon-Fri 14:30–21:00 UTC) |
BOOTSTRAP.md | First-run identity setup workflow |
HEARTBEAT.md | Periodic task template |
How to Use
First Run
Follow BOOTSTRAP.md to establish identity and connect with user.
Every Session
- Read
SOUL.md→USER.md - Load
memory/YYYY-MM-DD.md(today + yesterday) - In direct chat: also load
MEMORY.md
Stock Screener
The screener runs on GitHub Actions every hour during US trading hours. To regenerate manually:
python scripts/generate_screener.py
The output writes to stock_screener.html.
Memory Consolidation
# Session statistics
python memory_consolidation/memory_consolidation.py stats
# Generate short-term memory
python memory_consolidation/memory_consolidation.py stm
# Compact STM → LTM via LLM
python memory_consolidation/memory_consolidation.py compact
# Generate daily diary
python memory_consolidation/memory_consolidation.py diary
Notes
- Memory config lives at
memory_consolidation/memory_consolidation.env MEMORY.mdonly loads in direct 1:1 sessions — never in group contextsworker-safetyskill enforces hard limits: no runtime/plugin upgrades via agent, gateway stays local- Screener fetches from Yahoo Finance with S&P 500 and NASDAQ-100 as data sources
还没有评论。