← 返回 Skills 市场

HFT Paper Trader Pro

Name: HFT Paper Trader Pro
Author: zero2ai-hub

作者 Zero2Ai · GitHub ↗ · v1.1.0 · MIT-0

cross-platform ⚠ suspicious

225

总下载

当前安装

版本数

在 OpenClaw 中安装

/install hft-paper-trader-pro

功能描述

High-frequency paper trading framework for crypto. Multi-indicator TA scoring (RSI/MACD/EMA/BB/OBV/StochRSI), dual-regime filter (15m fast + 4h macro), posit...

使用说明 (SKILL.md)

HFT Paper Trader — Autonomous Crypto Trading Framework

A complete high-frequency paper trading system for building and testing autonomous crypto trading agents.

Architecture

Market Data (Binance public API)
    ↓
TA Engine (RSI + MACD + EMA + BB + OBV + StochRSI)
    ↓
Signal Score (-10 to +10)
    ↓
Dual Regime Filter (15m EMA8/21 fast + 4h EMA20/50 macro)
    ↓
Kelly Position Sizer (3% max risk per trade)
    ↓
Paper Portfolio Manager (portfolio.json)
    ↓
Trade Ledger (journal.json) + Auto-Observation Logger (observations.md)

Features

Multi-indicator confluence: 7 indicators combined into one score
Dual regime filter: 15m EMA8/21 (fast) gates entries alongside 4h EMA20/50 (macro) — prevents trading against short-term trend
OBV divergence detection: hidden accumulation/distribution
Quarter-Kelly sizing: conservative risk management
Correct SL placement: Math.max caps risk at 3% — fixes bug where SL ran 5–11% instead of 3%
Drawdown controls: auto-pause at 2% daily NAV
Full audit trail: every trade logged with entry/stop/target/outcome
Auto-observation logging: losses and SL hits automatically appended to observations.md for strategy learning
Self-improvement loop: lessons captured after each loss

Usage

Use hft-paper-trader to run TA on BTC and place a paper trade

Use hft-paper-trader to check portfolio performance

Use hft-paper-trader to scan the watchlist and trade all signals

Use hft-paper-trader to check today's observations and lessons

Regime Filter Logic

Entries are only allowed when BOTH regimes are bullish:

Fast (15m): EMA8 > EMA21
Macro (4h): EMA20 > EMA50

This gates out counter-trend entries that otherwise pass signal scoring. When either regime is bearish, new positions are blocked regardless of score.

Stop-Loss Placement (Fixed v1.1.0)

// CORRECT — caps SL at 3% below entry
stopLoss = entry * (1 - MAX_RISK);   // Math.max not used here — direct cap

// Bug in v1.0.0 — Math.min let SL run to 5-11%
// Fix: use direct percentage cap on entry price, not min of wick distance

Watchlist

BTC ETH SOL XRP TRX DOGE ADA AVAX BNB LINK LTC SUI ARB OP NEAR DOT ATOM UNI MATIC

File Layout

trading/
  paper-dashboard/portfolio.json   ← live portfolio state
  journal.json                     ← full trade ledger
  observations.md                  ← auto-logged trade lessons

Performance

Starting capital: $1,000. Runs hourly (XX:01 UTC). Max 5 concurrent positions at 10% each.

安全使用建议

Do not install or run this skill without additional verification. Key actions to take before trusting it: 1) Ask the publisher for the full source code or a canonical repository/homepage and confirm the owner identity (the registry metadata, SKILL.md, and _meta.json contain inconsistent owner/slug/version values). 2) Request explicit details about any external endpoints used by the 'self-improvement loop' (where lessons are sent/stored); disallow any unexpected network exfiltration. 3) Run the skill in an isolated environment (no access to sensitive network segments or credentials) and monitor outbound connections; ensure it only calls public Binance endpoints you expect. 4) Inspect any implementation for network calls, hardcoded URLs, or code that reads other credentials or system files. 5) Do not connect any real exchange accounts or real funds — this is described as paper trading. 6) If you need autonomous execution, prefer skills with a public repo, clear provenance, and reproducible builds; if the author can’t or won’t provide source/clarity, treat the skill as untrusted.

功能分析

Type: OpenClaw Skill Name: hft-paper-trader-pro Version: 1.1.0 The skill bundle contains metadata and instructions (SKILL.md) for an autonomous crypto paper trading agent. It outlines logic for technical analysis, risk management using the Kelly criterion, and trade logging. No executable code, suspicious network calls, or data exfiltration attempts were identified; the instructions are consistent with the stated purpose of simulating high-frequency trading using public Binance API data.

能力评估

ℹ Purpose & Capability

SKILL.md describes a coherent paper-trading framework (TA indicators, regime filters, file-based portfolio/ledger). That functionality does not require extra credentials or installers, so the declared lack of required env vars/install is plausible. However registry metadata (slug/version/owner) does not match _meta.json and SKILL.md (registry lists 'hft-paper-trader-pro' v1.1.0 while _meta.json and SKILL.md show different slugs/versions/owner), and there is no homepage or source URL — this provenance mismatch is unusual and unexplained.

ℹ Instruction Scope

The SKILL.md is instruction-only and describes reading market data from 'Binance public API' and writing local files (portfolio.json, journal.json, observations.md). Those actions are consistent with paper trading. But instructions are high-level and vague (no explicit endpoints, no commands, no code), and the 'self-improvement loop' / 'lessons captured after each loss' is underspecified (no target for telemetry or model training). Vague instructions can give an agent broad discretion to call external endpoints or aggregate and transmit data if implemented later.

✓ Install Mechanism

No install spec and no code files — lowest-risk packaging pattern. Nothing will be written to disk by an installer at install time. The security surface is entirely the SKILL.md instructions when the agent runs.

ℹ Credentials

The skill declares no required environment variables or credentials, which is reasonable for public-market-data-only paper trading. However, SKILL.md references 'Binance public API' (some Binance endpoints require keys/rate-limited access) and an autonomous/self-improvement loop that might need remote storage or credentials; the absence of any declared env vars or endpoints leaves ambiguity about what secrets (if any) might be needed or used by an implementation.

✓ Persistence & Privilege

Flags are default (always:false, user-invocable:true, model invocation allowed). The skill will not be force-included system-wide. It will be able to act autonomously if invoked, which is normal for skills; this combination alone is not a red flag.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install hft-paper-trader-pro
安装完成后，直接呼叫该 Skill 的名称或使用 /hft-paper-trader-pro 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.1.0

- Added a detailed SKILL.md with expanded architecture, feature set, and usage instructions. - Introduced a multi-indicator TA engine combining 7 indicators for signal scoring. - Implemented a dual-regime filter (15m EMA8/21 and 4h EMA20/50) to gate entries. - Fixed stop-loss logic: SL now hard-capped at 3% risk per trade (was 5–11% in v1.0.0). - Enhanced risk management with Kelly position sizing and daily NAV drawdown pause. - Improved trade journaling and automated observation logging for self-improvement.

元数据

Slug hft-paper-trader-pro

版本 1.1.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题