← Back to Skills Marketplace
lj22503

Investor Education Workflow

by lj22503 · GitHub ↗ · v1.2.0 · MIT-0
cross-platform ✓ Security Clean
80
Downloads
0
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install investor-education-workflow
Description
[何时使用]当用户需要投资者教育内容时;当用户说'写篇投教文章'、'解释这个投资概念'、'做个投教卡片'、'定投是什么'、'如何防骗'时触发。基于 LLM Wiki 机制的投教内容生产与分发工作流。
README (SKILL.md)

investor-education-workflow: 投教工作流 🎯

📋 功能描述

帮助用户系统化生产与分发投资者教育内容。基于 LLM Wiki 机制,实现"知识库优先 → 搜索补充 → 转译大白话 → 多形式输出 → 反馈入库"的完整闭环。

适用场景:

  • 投教内容生产(文章/卡片/语音/视频脚本)
  • 知识点转译(专业术语 → 大白话)
  • 客户问答响应(1 对 1/群发/讲座)
  • 知识库维护(LLM Wiki编译与更新)

边界条件:

  • 不替代持牌投教机构
  • 知识准确性与合规由 IE 负责,内容生成与格式转换由 expression-layer 负责
  • 知识库路径:knowledge/investor-education/wiki/
  • 原始素材路径:knowledge/investor-education/raw/

🔄 核心处理流程(5 阶段)

阶段①:检索知识库

动作:在 knowledge/investor-education/wiki/ 中全文检索关键词。 输出:命中 → 提取结构化知识点(定义/案例/话术);未命中 → 进入阶段②。 调用 SkillRead(读取 Wiki Markdown 文件)

阶段②:定向搜索补充 + 逻辑提取

触发条件:知识库未命中 / 内容过时 / 问题非常具体。 动作

  1. 搜索:限定官方域名(site:nerc.edu.cn 等),获取原始材料。
  2. 提取因果链:搜索内容往往是新闻摘要或碎片信息。必须先提取逻辑,再转译。
    • 识别核心变量(如:降准 → 利率 ↓ → 债基 ↑)
    • 构建因果链(A 导致 B,B 影响 C)
    • 剔除矛盾/过时信息 输出:结构化因果链 + 来源标注。 调用 Skillsearxng + url-to-markdown

阶段③:合规拦截 + 转译大白话 + 意图标记

动作

  1. 合规拦截:检测用户请求是否触碰红线。
    • 🔴 红线:推荐具体基金/股票代码、承诺收益、预测短期涨跌、代客理财。
    • 🟡 转化协议:若触碰红线,不直接拒绝,而是转化为投教内容。
      • 例:用户"推荐只下周必涨的基" → 转化为"如何自己筛选优质基金"或"为什么预测短期涨跌是陷阱"。
      • 标记 intent: plaincard,输出教育性内容。
  2. 转译:将阶段①/②的内容转译为大白话。
    • 去术语化:用生活类比("PE=回本年限,像买店铺看租金")
    • 场景化:嵌入普通人能遇到的情境
    • 行为化:直接告诉客户"你该做什么/不该做什么"
  3. 意图标记(必须输出结构化参数):
    intent: plain | writes | card | wechat | multi
    mold: -l | -i | -c | -w | -b  # 仅 intent=card 时必填
    audience: 儿童 | 新手 | 进阶 | 专业  # 可选,指导转译难度
    

输出:结构化内容 + 意图参数。 调用 Skillljg-learn(概念解剖)

阶段④:多形式输出(调用表达层)

动作:将阶段③的内容 + 意图参数传递给 expression-layer调用方式

调用 expression-layer,传入:
- content: [阶段③的结构化内容/因果链]
- intent: [plain/writes/card/wechat/multi]
- mold: [-l/-i/-c/-w/-b]  # 仅 card 时传
- audience: [儿童/新手/进阶/专业] # 可选

输出:表达层返回最终成品(Markdown/PNG/HTML/公众号)。 调用 Skillexpression-layer

阶段⑤:反馈入库

动作:将本次生成的内容、客户反馈、使用频次结构化回填至知识库。 入库路径

  • 新知识点 → knowledge/investor-education/wiki/entities/concepts/
  • 客户问答 → knowledge/investor-education/raw/qa/
  • 反馈标记 → 在 Wiki 页面底部追加元数据:
    ---
    使用频次:高/中/低
    理解难度:⭐/⭐⭐/⭐⭐⭐/⭐⭐⭐⭐/⭐⭐⭐⭐⭐
    最后更新:2026-04-24
    关联偏误:过度自信/损失厌恶/羊群效应
    ---
    

调用 Skilltask-state-tracker


📚 五大模块知识库

模块 内容 对应 Skill
模块 1:投资基本功 金融基础/基本面分析/技术分析/财经素养 ljg-learn, fund-analyzer-pro
模块 2:市场认知与风险扫描 资本市场全景/风险识别/合规交易 decision-checklist, content-compliance
模块 3:行为管理 行为金融学/常见偏误/策略矫正 decision-checklist, ljg-relationship
模块 4:投资规划与生命周期 财务规划/资产配置/退休规划 fund-allocator, ljg-rank
模块 5:投资者陪伴与持续成长 分层学习路径/多元形式/日常陪伴 companion-script, ljg-plain, ljg-card

详细五大模块说明 → references/five-modules.md LLM Wiki 架构 → references/llm-wiki.md 四专家思维框架 → references/four-experts.md


⚠️ 常见错误

错误 1:知识库不存在

问题:
• 阶段①检索失败,因为 knowledge/investor-education/ 目录未创建
• 流程断裂

解决:
✓ 首次使用时,先创建目录骨架(mkdir -p knowledge/investor-education/{raw,wiki/{entities,concepts,summaries}})
✓ 导入种子数据(10 个高频知识点 Wiki 页面)
✓ 若目录不存在,直接跳至阶段②搜索

错误 2:未调用表达层

问题:
• 阶段④直接输出 Markdown,未调用 expression-layer
• 无法生成卡片/公众号/语音

解决:
✓ 阶段④必须调用 expression-layer,传入 content + intent + mold
✓ 不要自己生成 PNG/HTML,让表达层路由

错误 3:意图标记缺失或参数不全

问题:
• 未标记 intent 或 mold,表达层不知道要出 plain 还是 card,或默认用错模具
• 路由失败或输出格式不匹配

解决:
✓ 阶段③必须输出完整意图参数(intent + mold + audience)
✓ 歧义时主动询问用户:"需要大白话解释、深度文章,还是可视化卡片?大字还是信息图?"

错误 4:搜索源非官方

问题:
• 阶段②搜索返回自媒体文章,非官方源
• 内容合规风险

解决:
✓ 搜索时必须限定 site:nerc.edu.cn OR site:sse.org.cn OR site:szse.cn
✓ 优先使用中国投资者网、交易所投教基地

错误 5:合规请求硬拒绝

问题:
• 用户问"推荐只下周必涨的基",AI 直接回复"我不能推荐"
• 错失教育机会,用户体验差

解决:
✓ 执行"合规拦截与转化"协议
✓ 不拒绝,而是转化:"预测短期涨跌是陷阱,我教你 3 个自己筛选基金的方法..."
✓ 标记 intent: plain 或 card,输出教育性内容

🧪 使用示例

输入:

客户问:"什么是市盈率(PE)?能不能说人话?"

预期输出:

  • 阶段①:检索知识库 → 命中"市盈率"Wiki 页面
  • 阶段③:转译大白话 → 标记 intent: plain
  • 阶段④:调用 expression-layer → 返回口语化解释(≤200 字)
  • 阶段⑤:入库 → 标记"理解难度⭐⭐",关联"锚定效应"

输入:

写篇投教文章:定投为什么能对抗择时焦虑?发到公众号。

预期输出:

  • 阶段①/②:检索/搜索 → 获取定投原理 + 行为金融学素材
  • 阶段③:转译 → 标记 intent: wechat
  • 阶段④:调用 expression-layer → 返回公众号推文(HTML+ 封面)
  • 阶段⑤:入库 → 标记"使用频次高"

🔧 故障排查

问题 检查项
阶段①失败 knowledge/investor-education/wiki/ 是否存在?是否有种子数据?
阶段④失败 是否调用了 expression-layer?intent 标记是否正确?
输出格式错误 expression-layer 路由矩阵是否包含该意图?
搜索非官方 搜索命令是否包含 site:nerc.edu.cn 等限定?
未入库 阶段⑤是否执行?wiki 页面底部元数据是否追加?

🔗 相关资源

  • 五大模块知识库:references/five-modules.md
  • LLM Wiki 架构:references/llm-wiki.md
  • 四专家思维框架:references/four-experts.md
  • 表达层路由:../expression-layer/SKILL.md
  • 报告模板:templates/education-template.md
  • 标准参考:docs/SKILL-STANDARD-v3.md
Usage Guidance
This skill appears internally coherent for producing investor-education content, but review these before installing: - Storage & privacy: The workflow saves customer Q&A and generated content into knowledge/investor-education/raw/ and wiki/ — if those records can contain personal data, confirm retention, redaction, and access controls. Consider seeding the knowledge directory in a controlled way before real data is used. - Downstream skills: The SKILL.md delegates formatting/publishing to expression-layer and uses search helpers (searxng, url-to-markdown). Audit those skills/endpoints to ensure they don't leak content to untrusted external services or require credentials you don't want to share. Verify how expression-layer publishes (e.g., WeChat/HTML/PNG) and what external APIs it calls. - Shell/Web capabilities: Allowed-tools include Exec/Bash/WebSearch. If your agent environment allows shell execution, confirm sandboxing limits so the skill cannot run arbitrary commands outside the knowledge paths. - Compliance behavior: The skill intentionally 'soft-transforms' red-line requests (e.g., 'recommend this fund') into educational content rather than hard refusal. Confirm that this approach meets your legal/regulatory obligations and that the transformation logic is implemented safely (no accidental giving of actionable investment advice). - Testing: Test the skill in a safe environment with non-sensitive example queries, verify the search-source restrictions are actually applied, validate intent/mold/audience parameter propagation to expression-layer, and inspect what gets persisted to the wiki before enabling on production data. If you cannot audit the downstream skills or control where generated content is published, treat the skill as higher risk and do not enable it for sensitive data or autonomous operation.
Capability Analysis
Type: OpenClaw Skill Name: investor-education-workflow Version: 1.2.0 The investor-education-workflow bundle is a legitimate tool for generating financial education content using a structured 'LLM Wiki' approach. It includes robust safety features, such as a compliance protocol in SKILL.md that explicitly instructs the agent to intercept requests for specific investment advice and convert them into educational topics. While the skill requests high-privilege tools (Bash, Exec), its instructions are limited to local file management within the 'knowledge/investor-education/' directory and performing targeted web searches on official Chinese financial regulatory domains (e.g., site:nerc.edu.cn). The presence of detailed development logs and architectural documentation (evolve-test/ and references/) further supports its legitimacy.
Capability Assessment
Purpose & Capability
Name/description state a content production/distribution workflow. The SKILL.md only requires reading/writing local knowledge paths, web search, and routing content to an 'expression-layer' — all coherent with the stated purpose. No unrelated environment variables, cloud credentials, or surprising binaries are requested.
Instruction Scope
Instructions explicitly read/write under knowledge/investor-education/*, perform web searches (with domain limits), extract logic/causal chains, call other skills (searxng, url-to-markdown, expression-layer, task-state-tracker), and persist generated Q&A back to the wiki/raw paths. That's consistent with a knowledge-workflow, but it means the skill will: (a) store user/customer Q&A (potential PII) in the knowledge directories; (b) send content to external routing/formatting components (expression-layer) which may publish to public channels. Confirm retention, redaction, and publishing policies for customer data.
Install Mechanism
Instruction-only skill with no install spec and no third-party downloads — lowest install risk. All runtime behavior is driven by SKILL.md and other skills; there is no archive download or execution artifact in this package.
Credentials
The skill requests no environment variables, no credentials, and no config paths beyond its knowledge directories. This is proportionate to a content production workflow. However, it relies on other skills (expression-layer, searxng, url-to-markdown) — those downstream skills may require credentials or external endpoints; review them before enabling.
Persistence & Privilege
always:false (normal). The skill writes to and expects a local knowledge directory structure (creates directories and appends metadata). That level of persistence is appropriate for a wiki-style workflow, but it grants the skill write access to local agent storage (knowledge/investor-education/*). Ensure the storage location is scoped and audited and that the skill cannot write outside its stated paths. Also confirm policies for autonomous invocation since the agent may run these steps without additional prompts.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install investor-education-workflow
  3. After installation, invoke the skill by name or use /investor-education-workflow
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.2.0
新增意图标记/知识库引用/因果链标注/合规检查记录
v1.0.0
v1.0.0: 基于 LLM Wiki 机制的投教工作流,融合四专家思维,覆盖五大模块知识库
Metadata
Slug investor-education-workflow
Version 1.2.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 2
Frequently Asked Questions

What is Investor Education Workflow?

[何时使用]当用户需要投资者教育内容时;当用户说'写篇投教文章'、'解释这个投资概念'、'做个投教卡片'、'定投是什么'、'如何防骗'时触发。基于 LLM Wiki 机制的投教内容生产与分发工作流。 It is an AI Agent Skill for Claude Code / OpenClaw, with 80 downloads so far.

How do I install Investor Education Workflow?

Run "/install investor-education-workflow" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Investor Education Workflow free?

Yes, Investor Education Workflow is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Investor Education Workflow support?

Investor Education Workflow is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Investor Education Workflow?

It is built and maintained by lj22503 (@lj22503); the current version is v1.2.0.

💬 Comments