mike
b660debd06
新增 8 reference: - dividend-stability-score: A 股 5 维评分 (派息年数/CAGR/波动/最近/连续) → 0-100 分 + 5 星 - dividend-yield-rate-sort: 按股息率% 倒序 (用户偏好 2026-07-13) - longport-http-module: longport_http.py 公共模块 (替代 SDK WSS) - leverage-pass-through-bug: process_signal.py 丢失 leverage 字段 (5x 实际 10x) - follow-trading-iron-laws: 跟单铁律 (用户原话 5+ 次 2026-07-21) - forced-skill-entry-okx-trade: okx_trade.sh 强制入口 (替代 ccxt 裸调) - mihomo-clash-node-supplier-dns: Clash 节点供应商 DNS 失败处理 - mihomo-ssl-reconnect-pattern: mihomo 反复 SSL/Timeout 模式 - v4.5.44-mu-add-to-75pct-cap: MU 加仓 75% 单币种 cap 标准流程 改 2 SKILL.md: - dividend-investing: 加 5 维评分 + 长桥 http 模块 - longbridge-cli: 标注 '不要写 openapi.QuoteContext' + 迁移说明
147 lines
5.3 KiB
Markdown
147 lines
5.3 KiB
Markdown
# OKX process_signal.py 杠杆丢失 Bug - 实战复现
|
||
|
||
**Captured**: 2026-07-13
|
||
**Skill version**: okx-auto-position v4.5.2+
|
||
**Severity**: Critical (风险放大 2-5 倍,爆仓概率翻倍)
|
||
|
||
## Symptom
|
||
|
||
`process_signal.py` 收到的信号里 `leverage` 字段(如 2x / 5x),advisor.execute 实际下单时**始终是 10x**——process_signal.py 把信号的 leverage 字段丢了,直接传默认值 10 给 advisor。
|
||
|
||
## Confirmed Cases (3 次)
|
||
|
||
### Case 1: 风寻 SKHY short (2026-07-08)
|
||
- 信号: 528.16 SKHY short @5x @161.90, +1.68% PnL
|
||
- 实际 execute: SKHY short 2.89张 @**10x** @avgPx 161.01
|
||
- 强平价: 192.86 (vs 信号 5x 应该 ~165,实际 10x 推到 192)
|
||
- 浮盈: +$1.88 (跟单成功,但杠杆翻倍 = 风险翻倍)
|
||
|
||
### Case 2: 风寻 SKHY short (2026-07-13)
|
||
- 信号: 3973.30 SKHY short @**2x** @157.30, +0.64% PnL
|
||
- 实际 execute: SKHY short 3.18张 @**10x** @avgPx 157.02
|
||
- 强平价: 187.95 (信号 2x 应该 ~157+(157/2)*0.01 = 157.78,实际 10x 推到 188)
|
||
- 浮盈: +$1.88
|
||
|
||
### Case 3: 熬鹰 MU short (2026-07-13)
|
||
- 信号: 664.37 MU short @**2x** @937.62, -0.01% PnL
|
||
- 实际 execute: MU short 0.52张 @**10x** @avgPx 940.40
|
||
- 强平价: 1136.92 (信号 2x 应该 ~940+(940/2)*0.01 = 944.70,实际 10x 推到 1137)
|
||
- 浮亏: -$0.28
|
||
|
||
### Case 4: 熬鹰 BTC long 20x (2026-07-21)
|
||
- 信号: 49.958 BTC long @20x @65547.73
|
||
- 实际 execute: 0.01 张 BTC long @10x @65407.6 (用 OKX 私有 API setLeverage 强制设 20x, 仍被覆盖成 10x)
|
||
- 浮亏: -0.01 USDT (立刻平了)
|
||
- **教训**: setLeverage 私有 API 不生效, 下单时仍用 10x (process_signal 内部写死)
|
||
- **新加 bug**: OKX 最小下单单位 min_sz 触发 → 我传 0.014 张, 实际成交 0.01 张 (0.01 是 min_sz)
|
||
- 双重 bug: leverage 5→10 + min_sz 截断。**两个都没在 process_signal 修过**
|
||
|
||
## 规律
|
||
|
||
- 信号 5x → 实际 10x (2 倍)
|
||
- 信号 2x → 实际 10x (5 倍)
|
||
- **execute 一律用默认值 10x,从不读信号里的 leverage**
|
||
|
||
## Root Cause (推测)
|
||
|
||
`process_signal.py` 调用 advisor 时,`leverage` 字段可能是:
|
||
1. 没传 → advisor 默认 10
|
||
2. 传了但被覆盖成 str(10)
|
||
3. parse_signal 的 leverage 字段解析错误(数字 + 'x' 后缀没去掉)
|
||
|
||
## Agent 侧强制校验流程
|
||
|
||
```python
|
||
import json, time, hmac, hashlib, base64, requests
|
||
|
||
def okx_get(p, params=None, t=15):
|
||
# ... 标准 raw REST GET 签名 ...
|
||
pass
|
||
|
||
# 1. execute 之后立即反查
|
||
signal_leverage = 5 # 信号原文
|
||
pos = okx_get('/api/v5/account/positions', {'instId': 'SKHY-USDT-SWAP'})
|
||
for p in pos.get('data', []):
|
||
actual_leverage = int(p.get('lever', 10))
|
||
if actual_leverage != signal_leverage:
|
||
# 杠杆不对!manual close + 重开
|
||
# 1. close current position
|
||
close_body = {
|
||
'instId': 'SKHY-USDT-SWAP',
|
||
'tdMode': 'cross',
|
||
'side': 'buy' if float(p['pos']) < 0 else 'sell',
|
||
'posSide': 'net',
|
||
'ordType': 'market',
|
||
'sz': str(abs(float(p['pos']))),
|
||
'reduceOnly': True,
|
||
}
|
||
okx_post('/api/v5/trade/order', close_body)
|
||
# 2. reopen with correct leverage
|
||
open_body = {
|
||
'instId': 'SKHY-USDT-SWAP',
|
||
'tdMode': 'cross',
|
||
'side': 'sell' if signal_side == 'short' else 'buy',
|
||
'posSide': 'net',
|
||
'ordType': 'market',
|
||
'sz': str(contracts),
|
||
'lever': str(signal_leverage), # 显式传 leverage
|
||
}
|
||
okx_post('/api/v5/trade/order', open_body)
|
||
```
|
||
|
||
## 永久修复 (待改 process_signal.py)
|
||
|
||
```python
|
||
# process_signal.py parse_signal() 函数
|
||
def parse_signal(text):
|
||
# ...
|
||
lev_match = re.search(r'【杠杆】\s*[:：]?\s*(\d+)\s*[xX]', text)
|
||
if lev_match:
|
||
fields['leverage'] = int(lev_match.group(1))
|
||
# 验证范围
|
||
if not 1 <= fields['leverage'] <= 50:
|
||
fields['leverage'] = 10 # fallback
|
||
return fields
|
||
|
||
# 然后调 advisor 时:
|
||
cmd = ['python3', 'okx_position_advisor.py', '--symbol', symbol,
|
||
'--side', side, '--leverage', str(fields['leverage'])] # 不用默认 10
|
||
```
|
||
|
||
## 已知受害币种
|
||
|
||
- SKHY (2 次)
|
||
- MU (1 次)
|
||
- 任何信号标 2x / 5x 的小币种永续合约都需校验
|
||
|
||
## 实战价值
|
||
|
||
- 5x → 10x:风险 2 倍,强平价远 50%
|
||
- 2x → 10x:风险 5 倍,强平价远 100%+
|
||
- 10x 杠杆下,1% 价格波动 = 10% 保证金波动,极容易爆
|
||
|
||
## 相关 SKILL.md 章节
|
||
|
||
- "v4.5.2 process_signal 杠杆丢失 bug" - 主入口
|
||
- "v4.5.0 平仓信号自动跟单" - 检测时需查 lever 字段
|
||
## Agent 行为铁律 (2026-07-17 用户原话)
|
||
|
||
### 1. 跟单不许反问
|
||
- **信号来了立刻 advisor → 下单**
|
||
- 不问 yes/no/几张
|
||
- 用户说"按信号来"=立即执行,不讨论
|
||
- (累计 3 次用户指出 2026-07-17 违反此规则)
|
||
|
||
### 2. 下单后立即验证 (不要把 advisor 推荐当事实)
|
||
- 下单后**立刻** `fetch_positions()` 查**真实**:lever, margin, notional, entryPrice
|
||
- advisor 推荐值 ≠ 实际成交值
|
||
- 实战教训(2026-07-17): 用户说"5x 杠杆", advisor 算 5x, 实际下单 10x (process_signal.py bug)
|
||
- 实际占 $26.92, 报告时错说 $53,**用户立即指出"又是猜的"**
|
||
- 教训:**下单后必须 fetch_positions 验证 3 件事**
|
||
- actual_leverage == requested
|
||
- actual_margin == expected
|
||
- actual_side == expected
|
||
|
||
## references index 更新
|
||
此 reference 是 OKX advisor 杠杆 + 下单事实校验的权威来源。任何新错误 / 修复补这里, MEMORY 只存指针。
|