菜单
Operate and reproduce the user's L2 ETF strategy package. Use when the user asks about L2 ETF, TRD_K20_ADV50M_NoKCB, quarterly ETF rebalancing, ETF picks/trades, live ETF holdings, NAV logs, or reproducing the ETF backtest/experiment evidence from the 2026-06-10 package.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | l2-etf-strategy |
| description | Operate and reproduce the user's L2 ETF strategy package. Use when the user asks about L2 ETF, TRD_K20_ADV50M_NoKCB, quarterly ETF rebalancing, ETF picks/trades, live ETF holdings, NAV logs, or reproducing the ETF backtest/experiment evidence from the 2026-06-10 package. |
Use this skill for the user's locked L2 ETF strategy line only. Do not mix it with Turtle stock rules, VCP experiments, or single-stock live trading logic.
Core asset root:
SKILL_DIR="${CODEX_HOME:-$HOME/.codex}/skills/l2-etf-strategy"
PKG="$SKILL_DIR/assets/L2_ETF_STRATEGY_EXPERIMENT_PACKAGE_2026-06-10"
LIVE="$PKG/01_live_strategy/L2_ETF_LIVE"
BACKTEST="$PKG/02_backtest_reproduction/L2_ETF_BACKTEST"
SOURCE_DATA="$PKG/03_source_data/L2_ETF_DATA_2026-05-29"
Always start by reading references/handoff.md. For live rebalancing or holdings work, also read references/live-strategy.md and references/live-changelog.md. For performance reproduction, read references/backtest-reproduction.md.
TRD_K20_ADV50M_NoKCB.588 or 589.The package cache is historical: source data is dated 2026-05-29. If a user asks for latest/current/future rebalance output, refresh or obtain up-to-date ETF pool and kline cache first, then run the same workflow with L2_ETF_DATA_ROOT pointing at the refreshed data. Do not present stale 2026-05-29 output as a current signal.
Use this when generating quarterly picks_YYYY-MM-DD.csv and trades_YYYY-MM-DD.csv.
current_holdings.json in LIVE and note confirmed live state. As of the package handoff, 2026-06-10 status is initial_build_completed_20_etfs; 159381 is intentionally above original equal-weight target and should be corrected only at the next quarterly rebalance unless the user explicitly requests interim action.SOURCE_DATA. For any current rebalance, use refreshed data with the same layout: all_etf_pool.csv plus kline_cache_all/*.json.cd "$LIVE"
L2_ETF_DATA_ROOT="$SOURCE_DATA" python3 quarterly_picks.py --cutoff YYYY-MM-DD --capital 50000
If the cutoff should be the newest common date in the cache, omit --cutoff. Use --show-pool-size for a quick eligibility check. Do not use --commit unless the user confirms actual fills or explicitly asks to update holdings.
Generated files are written beside quarterly_picks.py. When preserving canonical assets matters, copy LIVE to a temporary working directory first and run there with L2_ETF_LIVE_HOME set to that copy.
Use this to verify the locked performance evidence:
cd "$BACKTEST"
python3 run_capital_sweep.py
diff ./baselines_capsweep/capsweep_summary.csv ./capsweep_summary.csv
Expected headline in stdout:
TRD_K20_ADV50M ret: 1w:+110.5% 2w:+117.8% 3w:+119.0% 5w:+120.8% 10w:+122.6% 20w:+123.3%
No diff output means the regenerated capsweep_summary.csv matches the reference. The 50,000 RMB line corresponds to about +120.83% total return, -20.62% max drawdown, and 5.86 Calmar for 2024-01 to 2026-05.
The scripts require pandas and numpy. If the default python3 lacks them, use a Python environment that has both installed.
独立主控底盘型成长股正式研究；用私有 modules、多角色/subagent、中间产物、进攻型市场重定价主线、行业经济性与蛋糕分配、竞争/客户验证链、中期进攻 QA、终审事实 QA 和投资人写作，完整覆盖旧业务底盘、第二成长曲线、TAM/SAM/SOM、平台复用、承接动作、利润桥、跟踪体系、证据边界、三表/FCF 建模接力输入和估值接力输入。Use when 用户需要深度分析 A 股制造业、AI算力、数据中心、新能源、机器人、先进制造、半导体设备/材料等底盘型成长公司，判断旧业务托底、行业经济性拐点、新业务抬天花板、公司份额提升、平台能力复用、竞争客户验证、利润中枢上移、预期差和证伪路径，并产出正式深度报告而非摘要。
独立主控供应链平台型成长股正式研究；用私有 modules、多角色/subagent、中间产物、进攻型市场交易变量、技术路线到价值量、原材料价格链、竞争/客户认证链、产能升级、订单经营验证、利润天花板、跟踪体系、两轮 QA 和投资人写作，完整覆盖 AI 算力、数据中心、新能源、机器人、先进封装、半导体设备/材料等供应链公司研究。Use when 用户需要深度分析供应链公司，判断身份切换、订单放量、技术路线升级、利润中枢上移、天花板抬升、市场预期差和证伪路径，并产出正式深度报告而非摘要。
Real DCF (Discounted Cash Flow) model creation for equity valuation. Retrieves financial data from SEC filings and analyst reports, builds comprehensive cash flow projections with proper WACC calculations, performs sensitivity analysis, and outputs professional Excel models with executive summaries. Use when users need to value a company using DCF methodology, request intrinsic value analysis, or ask for detailed financial modeling with growth projections and terminal value calculations.
DCF-only valuation workflow orchestrator for equity research. Use when the user asks for DCF估值, DCF闭环, intrinsic value, Formal DCF, Scenario DCF, Reverse DCF, or wants a one-pass workflow that first prepares financial-modeling DCF-ready inputs and then runs dcf-model outputs. This skill coordinates financial-modeling to dcf-model for DCF deliverables only.
Build integrated financial models with 3-statement projections, DCF-ready UFCF bridges, working-capital schedules, and debt/interest linkages. Use for income statement, balance sheet, cash flow, and DCF input preparation. Scope is three-statement and DCF-ready modeling; PEG-ready packages are handled by growth-stock-valuation.
成长股 PEG/动态 PE 估值 skill；用于成长股、供应链平台股、产业链龙头的 PEG、动态 PE、复合 PEG、目标市值、目标价、一致预期差、当前市值隐含预期、估值年份切换和证伪点分析。目标市值默认使用自有正常/乐观利润锚，一致预期只作市场对照；范围限定为 PEG/动态 PE 成长定价。若用户要求 bottom-up/算小账/分业务利润桥与 PEG 并列对照，使用共享 catalyst precheck、独立 PEG subagent、独立 bottom-up subagent 和并列模块，防止估值与利润计算互相污染。