← Back to Skills Marketplace
118
Downloads
0
Stars
0
Active Installs
4
Versions
Install in OpenClaw
/install ausub
Description
基于Tushare黄金ETF日线行情和规则,动态生成带风险提示的黄金定投建议及投入金额参考。
README (SKILL.md)
OpenClaw
openclaw 是一个本地 skill,用来基于 Tushare 跟踪黄金行情,并输出动态定投建议。
适用场景
当用户希望做这些事情时,使用这个 skill:
- 基于 Tushare 获取黄金 ETF 的日度行情快照
- 用纪律化规则生成黄金动态定投建议
- 按固定节奏持续监测,而不是凭感觉择时
- 根据收入、支出、存款和投资偏好,生成一个初始黄金定投方案
输入
这个 skill 最好具备以下输入:
- 环境变量中的
TUSHARE_TOKEN - 要跟踪的标的定义
- 用户的基础定投金额,或者由模型协助生成的方案
- 可选的风控上下限
安装前说明
TUSHARE_TOKEN是必需项;如果当前运行环境没有正确设置这个变量,skill 将无法获取 Tushare 数据。- 安装其他 Tushare 相关 skill,并不等于当前 skill 会自动获得可用的
TUSHARE_TOKEN。 OPENCLAW_CONFIG、OPENCLAW_USER_ID、OPENCLAW_HISTORY_PATH、OPENCLAW_SYMBOL都是可选项,但如果设置了这些变量,skill 会读取对应的本地文件或路径。OPENCLAW_USER_ID会映射到memory/users/\x3Cid>.json。fetch_history.py会在references/history/下写入本地 CSV 文件。- 远程网络请求仅通过
tushare库访问 Tushare,不包含额外隐藏网络端点。
建议配置如下:
OPENCLAW_SYMBOL:跟踪的黄金 ETF,默认518880.SHOPENCLAW_BASE_AMOUNT:每期基础定投金额,默认推荐值可由用户覆盖OPENCLAW_MAX_MULTIPLIER:最大加码倍率,默认1.8OPENCLAW_MIN_MULTIPLIER:最小投入倍率,默认0.6
工作流
- 读取 Tushare 中配置标的的最新日线数据。
默认执行标的是
518880.SH。 ETF 主行情使用 Tusharefund_daily。 - 计算价格行为相关指标。
- 用核心指标把市场分成 4 个状态:
oversoldweakneutralhot
- 用布林线位置做倍率修正。
- 只在极端状态下使用 MACD 做确认修正。
- 如需展示黄金现货参考行情,读取上海黄金交易所
Au99.99。Au99.99参考行情使用 Tusharesge_daily。 - 输出带理由和风险提示的定投建议。
用户方案原则
weekly_thursday_if_trading和1000元应被视为默认推荐方案,而不是唯一固定方案。- 用户可以自行定义定投节奏、基础金额和执行偏好。
- 如果用户尚未确定方案,可以选择性描述预算范围、投资目标和风险偏好,由 openclaw 的模型给出一个建议方案。
- 不应要求用户在配置文件中持久化保存收入、支出、存款等敏感财务信息;这类信息仅适合在临时对话中用于生成建议。
- 涉及个人财务规划的方案建议,应在用户确认后再写入配置文件。
约束
- 不要把结果表述成保证收益的投资建议。
- 不要建议使用杠杆。
- 所有建议都要限制在设定的最小和最大倍率之间。
- 如果数据缺失、过旧或接口失败,要明确说明,不要强行给建议。
输出格式
输出应包含:
- 市场摘要
- 指标摘要
- 建议倍率
- 建议投入金额
- 触发细节
- 自然语言解释
- 风险提示
规则说明
- 使用硬阈值,避免边界条件重叠。
- 60 日回撤固定定义为:
(rolling_60d_high - close) / rolling_60d_high - 布林线修正使用位置值
pos = (close - lower_band) / (upper_band - lower_band):pos \x3C= 0:+0.20 \x3C pos \x3C= 0.25:+0.10.25 \x3C pos \x3C 0.75:00.75 \x3C= pos \x3C 1:-0.1pos >= 1:-0.2
- MACD 只用于
oversold和hot两种极端状态,并要求柱体动能连续两期收敛后才触发修正。
示例
如果基础金额是 1000,最终倍率是 1.5,则建议本期投入 1500,并附带简要说明,例如市场偏弱、价格接近布林下轨、下跌动能减弱等。
实现说明
- 优先使用小型 Python 脚本负责 Tushare 取数和指标计算。
- 所有阈值都应易于调参。
- 数据获取和建议逻辑要分离,便于后续替换数据源或升级规则。
- 以 ETF 日线为主要决策依据。
518880.SH主行情默认通过 Tusharefund_daily获取。- 上海黄金交易所
Au99.99可作为辅助参考锚,默认通过 Tusharesge_daily获取,但不替代执行标的。 - 不要把“当前脚本未取到 Au99.99 数据”表述成“Tushare 不支持 SGE 数据”;正确表述应是“当前参考数据暂不可用或当前流程未启用该接口”。
Usage Guidance
Before installing: (1) expect to provide a TUSHARE_TOKEN — the skill will fail without it; the registry incorrectly lists no required env vars, so update your expectations accordingly. (2) The skill may read a user memory file memory/users/<OPENCLAW_USER_ID>.json and will write/read CSVs under references/history/ — ensure those files are stored only in a trusted project directory. (3) The code only contacts Tushare via the tushare library (no hidden endpoints), but review any token you provide and avoid reusing high-privilege credentials. (4) If you need to verify safety, inspect the included scripts (openclaw.py, fetch_history.py, backtest.py) yourself or run them in a sandboxed environment with a limited-scope Tushare token. (5) The metadata mismatch (missing TUSHARE_TOKEN declaration) is likely sloppy packaging rather than malicious, but treat it as a red flag and prefer to review code before granting environment variables.
Capability Analysis
Type: OpenClaw Skill
Name: ausub
Version: 1.0.3
The skill bundle is a legitimate financial analysis tool designed to provide gold investment recommendations using the Tushare API. The Python scripts (openclaw.py, backtest.py, fetch_history.py) implement sound security practices, such as path traversal protection via the ensure_within_project function and input sanitization for symbols and user IDs. No evidence of data exfiltration, malicious execution, or prompt injection was found; the code strictly interacts with the Tushare service as described in the documentation.
Capability Assessment
Purpose & Capability
The skill's name/description (gold ETF dynamic DCA recommendations) aligns with the code and SKILL.md: it uses Tushare to fetch ETF and SGE data and computes indicators. However, the registry metadata claims no required env vars while both SKILL.md and all scripts require a TUSHARE_TOKEN; this mismatch is an incoherence between declared requirements and actual capabilities.
Instruction Scope
Runtime instructions and included Python scripts stay within the stated purpose: they call the tushare API, compute indicators, produce recommendations, and optionally read local user memory (memory/users/*.json) and write history CSV under references/history/. No hidden network endpoints or broad system data collection are present in the SKILL.md or scripts.
Install Mechanism
There is no install spec; the package is instruction-and-script-only. requirements.txt lists only 'tushare' which is appropriate for the stated purpose. No remote download URLs or archive extraction are used.
Credentials
The code requires the TUSHARE_TOKEN environment variable to access Tushare (appropriate for the purpose) but the skill metadata did not declare this requirement. The scripts also honor optional env vars (OPENCLAW_CONFIG, OPENCLAW_USER_ID, OPENCLAW_HISTORY_PATH, OPENCLAW_SYMBOL) which allow reading local project files (memory/users/<id>.json and references/history/*.csv). Those local reads/writes are reasonable for this skill but should be documented in registry metadata so users know what to provide and what files may be accessed.
Persistence & Privilege
The skill does not request always:true or system-wide privileges. It reads/writes only project-local files (references/history and memory/users) and does not modify other skills or global agent settings. Path-escape protections are present to restrict file access to the project directory.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install ausub - After installation, invoke the skill by name or use
/ausub - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.3
- Clarified that users should not be required to save sensitive financial details (income, expenses, deposits) in configuration files; these should only be used transiently in dialog to generate suggestions.
- Updated the default plan recommendation: users may now optionally describe budget range, investment goals, and risk preferences instead of fixed input requirements.
- Other instructions, constraints, and workflow details were unchanged.
v1.0.2
- 增加了“安装前说明”板块,详细解释了 Tushare Token 与本地配置文件等环境变量的作用和限制。
- 明确指出 skill 的远程网络请求只通过 tushare 库,不包含额外隐藏端点。
- 说明用户 id、本地历史数据等参数的自动映射和文件存储逻辑。
- 原有功能、规则和建议未作更改,核心使用方式保持不变。
- 微调部分措辞,进一步强调故障和缺失数据的处理细节。
v1.0.1
ausub 1.0.1
- No file changes detected in this version.
- No updates to SKILL.md or core functionality.
v1.0.0
AUSUB(OpenClaw)skill 初始版本发布:
基于 Tushare 数据,为黄金 ETF 提供自动化、规则化的动态定投建议。
跟踪指定黄金 ETF 的日度行情表现,默认标的为 518880.SH,并分析关键技术指标。
将市场划分为 oversold、weak、neutral、hot 四种状态,并结合布林线进行倍率修正,在极端状态下再用 MACD 做确认。
输出带有上下限约束的定投建议,同时提供清晰的触发原因说明和风险提示。
支持根据用户自定义的定投偏好与财务情况进行方案调整。
在数据缺失或数据过旧时,会明确说明情况,不会强行给出建议。
Metadata
Frequently Asked Questions
What is AUSUB?
基于Tushare黄金ETF日线行情和规则,动态生成带风险提示的黄金定投建议及投入金额参考。 It is an AI Agent Skill for Claude Code / OpenClaw, with 118 downloads so far.
How do I install AUSUB?
Run "/install ausub" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is AUSUB free?
Yes, AUSUB is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does AUSUB support?
AUSUB is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created AUSUB?
It is built and maintained by juryory (@juryory); the current version is v1.0.3.
More Skills