magiccube/agentara/stock

Stock Quote Skill

Fetch latest stock price, change, and anomaly detection for US stocks, HK stocks, and A-shares. Uses stooq.com as data source — reliable, no auth, bypasses system proxy automatically. Data is T+1 for all markets (previous trading day close). This is expected and normal.

Step 1 — Identify Stock & Market

Parse the user's query:

• A-share: 6-digit code (000001, 600519, 300750) or Chinese company name → market = cn
• HK stock: 4–5 digit code (9988, 00700, 9999) or XXXX.HK format → market = hk
• US stock: Latin ticker (BABA, AAPL, TSLA, NVDA) or BABA.US format → market = us

For well-known companies, infer the market:

• 阿里巴巴 → BABA (US) + 09988 (HK) — user usually means one, ask if unclear
• 腾讯 → 00700 (HK)
• 茅台/贵州茅台 → 600519 (A-share)
• 宁德时代 → 300750 (A-share)

For HK symbols, pass as-is (no zero-padding): 9988, 700. stooq handles it.

Step 2 — Fetch Data

Use the unified fetch_stock.py script. Market codes: hk, cn, us.

# HK stock
python3 /Users/henry/.agentara/.claude/skills/stock/fetch_stock.py hk 9988
 
# A-share
python3 /Users/henry/.agentara/.claude/skills/stock/fetch_stock.py cn 600519
 
# US stock
python3 /Users/henry/.agentara/.claude/skills/stock/fetch_stock.py us BABA

Output is a JSON array of rows: {date, open, close, high, low, vol, pct, chg}. If the output starts with {"error":, report the error to the user and stop.

Network notes: stooq.com is called with trust_env=False (bypasses Clash/system proxy). No external Python deps beyond requests (stdlib-equivalent, always available).

Stooq symbol format:

• HK: 9988.hk (no leading zeros)
• CN: 600519.cn
• US: baba.us (lowercase)

Step 3 — Analyze & Detect Anomalies

From the JSON output (up to 21 rows):

1. Latest row = most recent trading day data
2. Latest price = close of last row
3. Latest change% = pct of last row
4. Avg volatility = mean of |pct| of rows 2–21 (excluding last row)
5. Avg volume = mean of vol of rows 2–6 (5-day avg, excluding last)
6. Latest volume = vol of last row

Anomaly flags (report any that apply):

• |pct| > 5% → 🚨 大幅波动
• |pct| > 2% AND |pct| > 2 × avg_volatility → ⚠️ 异常波动（超出近期均值2倍）
• latest_vol > 2 × avg_volume → 📊 成交量异常放大（X倍）
• latest_vol < 0.4 × avg_volume → 🔇 成交量异常萎缩

If US stock pct is always near 0 except the last row, it's because the diff is computed on fetched data — use the last row's pct directly.

Step 4 — Format Output

Keep it brief and mobile-friendly. Use list format, no tables.

Color + emoji rules (Chinese convention: 红涨绿跌):

• pct > 0 → emoji 📈, wrap numbers in <font color='red'>...</font>
• pct < 0 → emoji 📉, wrap numbers in <font color='green'>...</font>
• pct == 0 → emoji ➡️, no color wrapper

Template (adapt to context):

**{公司名} · {市场} {代码}**
 
- 最新价：<font color='red/green'>**{price} {货币}**</font>
- 涨跌：{emoji} <font color='red/green'>**{chg:+.2f} / {pct:+.2f}%**</font>
- 最新交易日：{date}
 
{anomaly_section_if_any}

Anomaly section (only if flags exist):

⚠️ 异常提示
- [flag 1]
- [flag 2]
近5日均幅 **{avg_vol:.2f}%**，今日 <font color='red/green'>**{pct:.2f}%**</font>

If no anomaly: omit the anomaly section entirely. Keep the whole response under 10 lines.

Step 5 — Currency

• A-share: ¥ (人民币)
• HK: HK$ (港元)
• US: $ (美元)

Notes

• US data lags by 1 trading day (Sina source) — this is expected, tell the user if they ask
• If AKShare is not installed: pip3 install akshare -q
• If the symbol is wrong or not found, tell the user and suggest the correct format
• For A-share names, you can infer the 6-digit code from common knowledge; if uncertain, say so
• Do NOT show raw JSON or DataFrame output to the user — always parse and format it