hermes-closing-snapshot

name	hermes-closing-snapshot
description	Execute the Hermes A-share theme-structure SOP closing snapshot pipeline. Use when the user asks to run the closing snapshot, generate post-market sentiment/theme/performance outputs, or update hermes/outputs/ with today's closing data.
agent_created	true

Hermes A股题材结构SOP收盘快照

When to use

• 用户要求执行 "A股题材结构SOP收盘快照"
• 需要生成 sentiment_YYYY-MM-DD.json、theme_hypothesis_YYYY-MM-DD.json、performance_YYYY-MM-DD.json
• 数据源为 westock-mcp / westock-data，输出目录为 hermes/outputs/

Workflow

1. Fetch closing market data via westock-mcp:
   • mcp__westock-mcp__data_quote for sh000001,sz399001,sz399006
   • mcp__westock-mcp__data_kline for the same indices (day, limit 20)
   • mcp__westock-mcp__data_changedist with {"type":"0"} for up/down/limit counts
   • mcp__westock-mcp__data_market_overview with type=trade,updown,summary
   • mcp__westock-mcp__data_sector with mode=ranking,limit=50 for sector gainers & fund flow
2. Write hermes/data/market_snapshot.json with closing fields:
   • date, session: "close", index_changes, index_closes, index_kline_close
   • up_count, down_count, flat_count, up_down_ratio
   • limit_up_count, limit_down_count, volume_ratio, total_amount_yi, amount_change_yi
   • northbound_flow (set to 0 if no live field available)
   • consecutive_days.limit_up_trend = "rising"/"falling" vs previous session
   • hot_boards, sector_data, limit_pool_close, stocks: []
3. Ensure hermes/data/candidate_history.json exists (create [] if missing).
4. Run the three SOP scripts with --date YYYY-MM-DD:
   • hermes/scripts/sentiment_snapshot.py
   • hermes/scripts/theme_hypothesis.py
   • hermes/scripts/performance_tracker.py
5. Verify outputs in hermes/outputs/ and present them to the user.
6. Append a brief note to the workspace daily log (WorkBuddy/.workbuddy/memory/YYYY-MM-DD.md).

Notes

• research_only boundary: do not emit actionable buy/sell advice; only analysis, regimes, and scoring.
• Discord notifications in the scripts fail silently if DISCORD_WEBHOOK_URL is unset; this is expected.
• If candidate_history.json is empty, performance_tracker.py will report 0 candidates.