القائمة
通过兼容 SKILL.md 协议的第三方智能体查询 GemStar 量化交易提醒、最新信号、目标持仓和运行状态；只允许查看或显式触发一次本地研究流程，禁止实盘下单。
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
| name | gemstar-skill |
| description | 通过兼容 SKILL.md 协议的第三方智能体查询 GemStar 量化交易提醒、最新信号、目标持仓和运行状态；只允许查看或显式触发一次本地研究流程，禁止实盘下单。 |
Use this skill when the user asks from a compatible skill host about GemStar, current holdings, target holdings, trading alerts, live signals, paper ledger, leaderboard, bundled strategies, or whether GemStar has buy/sell guidance.
GemStar lives at:
/Users/ken/workspace/GemStar
Run all GemStar commands from that directory.
Current holdings, target holdings, and rebalance actions:
cd /Users/ken/workspace/GemStar
test -f artifacts/current/trade_status.md && cat artifacts/current/trade_status.md
Use the bundled script when the current directory is unknown:
bash /Users/ken/workspace/GemStar/integrations/gemstar-skill/scripts/gemstar-status.sh
Machine-readable current trade status:
cd /Users/ken/workspace/GemStar
test -f artifacts/current/trade_status.json && cat artifacts/current/trade_status.json
Latest chat-friendly alerts:
cd /Users/ken/workspace/GemStar
uv run python -m src.cli.app alerts latest --limit 5
Use the bundled script when the current directory is unknown:
bash /Users/ken/workspace/GemStar/integrations/gemstar-skill/scripts/gemstar-alerts.sh 5
Machine-readable latest alerts:
cd /Users/ken/workspace/GemStar
uv run python -m src.cli.app --output json alerts latest --limit 5
Latest leaderboard:
cd /Users/ken/workspace/GemStar
uv run python -m src.cli.app leaderboard
Bundled leaderboard strategy:
cd /Users/ken/workspace/GemStar
test -f strategies/leaderboard_quality_lowvol_v1/config.yaml && cat strategies/leaderboard_quality_lowvol_v1/config.yaml
Pipeline status:
cd /Users/ken/workspace/GemStar
uv run python -m src.cli.app status
Only run this when the user explicitly asks to refresh or run GemStar now:
cd /Users/ken/workspace/GemStar
uv run python -m src.cli.app trade --once
This may take several minutes because it can run the deterministic production pipeline, target generation, and one live cycle. It does not invoke LLM research unless the user explicitly runs gemstar research.
artifacts/current/trade_status.json first; fall back to trade_status.md, then alerts.ref_date and file updated_at first. If ref_date is older than today's date, explicitly say the current completed GemStar status is stale and do not present it as today's latest signal.uv run python -m src.cli.app leaderboard and summarize rank, strategy, status, Sharpe, CAGR, max drawdown, and alpha.strategies/leaderboard_quality_lowvol_v1/config.yaml and summarize universe, timer, factor weights, top_n, rebalance frequency, and backtest window.300750.SZ 宁德时代.artifacts/current/trade_status.md yet and suggest running gemstar trade --once.