← 返回 Skills 市场
FearHarvester
作者
bowen31337
· GitHub ↗
· v0.1.0
480
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install fear-harvester
功能描述
Automates dollar-cost averaging into BTC/ETH during extreme fear (F&G <10) and rebalances to yield above F&G 50, removing emotion from fear buying.
使用说明 (SKILL.md)
FearHarvester Skill
Autonomous DCA agent for extreme fear markets.
Strategy
- Monitor Fear & Greed index continuously
- When F&G \x3C 10 (Extreme Fear): DCA into BTC/ETH
- When F&G > 50 (Neutral/Greed recovery): rebalance into yield
- Removes human emotion from "buy the fear"
Usage
uv run python scripts/backtest.py --start 2018-01-01 --capital 10000
uv run python scripts/signals.py --live
uv run python scripts/executor.py --dry-run
Historical edge (2018-2024)
Buying F&G \x3C 10, holding 90d → 40-80% average return
安全使用建议
This skill generally does what it says (fetches Fear & Greed and Binance prices, backtests, and records DCA buys), but it contains a few red flags you should consider before installing or running it:
- 'Live' mode is misleading: executor.py prints that live requires API keys, but there is no code to place orders on an exchange or to read/store API keys. If you expect automatic trading, this skill will not perform real exchange orders as-is. Do not assume funds will be moved — the script only updates local state.
- Hard-coded absolute path in backtest.py: it writes results to /home/bowen/..., which may fail or overwrite files on your system. Expect to edit that path before running or run in an isolated environment.
- No declared credential handling: if you modify this to integrate real execution, you must add secure credential handling (do NOT paste API keys into the script or commit them).
- Network I/O: the scripts call external services (api.alternative.me and api.binance.com). That is expected, but be aware of network traffic and possible rate limits.
- Source trustworthiness: no homepage and unknown source. Prefer to run only after reviewing and/or auditing the code, run in --dry-run or --paper mode first, and execute in an isolated environment (container or sandbox). If you plan to enable actual trading, implement vetted exchange integration, secrets management, and thorough testing before connecting real funds.
功能分析
Type: OpenClaw Skill
Name: fear-harvester
Version: 0.1.0
The OpenClaw skill 'FearHarvester' is benign. It implements a cryptocurrency trading strategy by monitoring the Fear & Greed index and BTC prices from legitimate public APIs (alternative.me, Binance). The `SKILL.md` contains no prompt injection attempts. The Python scripts (`backtest.py`, `executor.py`, `signals.py`) perform their stated functions without attempting to exfiltrate data, execute arbitrary commands, or establish persistence. The `executor.py` script correctly uses relative paths for its state file, and while `backtest.py` uses a hardcoded absolute path (`/home/bowen/clawd/skills/fear-harvester/data/backtest_results.json`) for saving results, this is a minor portability flaw rather than a security vulnerability or malicious act, as it targets a data file within the skill's presumed installation directory.
能力评估
Purpose & Capability
Name/description align with the code: scripts fetch Fear & Greed and Binance prices and implement DCA/backtest/rebalance logic. However the executor advertises a 'live' mode that 'requires API keys' yet the code contains no exchange-ordering logic nor any credential handling—it only records state locally. That mismatch between claimed capability (live trading) and actual implementation is unexpected.
Instruction Scope
SKILL.md tells users to run the backtest, signals, and executor (including --live). The scripts perform network calls to third-party endpoints (api.alternative.me and api.binance.com) which is expected. Concerns: backtest explicitly writes results to an absolute path (/home/bowen/...), which can overwrite or fail depending on the environment; executor --live will update local state (and stop treating runs as dry-run) without requiring or verifying real exchange API keys, which could mislead a user into thinking trades will be executed when they are only being recorded locally.
Install Mechanism
No install spec; code is included in the bundle and requires only Python and requests. No external archives or arbitrary downloads are performed by an installer. This is the lower-risk install model, though running the scripts will perform network I/O at runtime.
Credentials
The skill declares no required environment variables or credentials, but the executor mentions 'live (requires API keys)'. There is no code to read API keys or other secrets, which is inconsistent: a legitimate live-execution skill would request and use exchange API keys. Right now there is no credential exfiltration, but the mismatch is misleading and could cause a user to later paste keys into an ad-hoc location or modify the code insecurely.
Persistence & Privilege
always:false (normal). The skill writes persistent state to data/executor_state.json (relative) and the backtest writes to a hard-coded absolute path in /home/bowen/..., which is poor hygiene. It does not alter other skills or system configurations. Autonomous invocation is allowed by default (not flagged here), but combined with the 'live' mislabeling and state writes, users should be careful about unattended runs.
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 fear-harvester - 安装完成后,直接呼叫该 Skill 的名称或使用
/fear-harvester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
v0.1.0
Initial release: autonomous DCA agent for extreme fear markets. Backtested Sharpe 2.01. Monitors F&G index, executes systematic buys below threshold 10, rebalances at recovery.
元数据
常见问题
FearHarvester 是什么?
Automates dollar-cost averaging into BTC/ETH during extreme fear (F&G <10) and rebalances to yield above F&G 50, removing emotion from fear buying. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 480 次。
如何安装 FearHarvester?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install fear-harvester」即可一键安装,无需额外配置。
FearHarvester 是免费的吗?
是的,FearHarvester 完全免费(开源免费),可自由下载、安装和使用。
FearHarvester 支持哪些平台?
FearHarvester 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 FearHarvester?
由 bowen31337(@bowen31337)开发并维护,当前版本 v0.1.0。
推荐 Skills