← Back to Skills Marketplace
99
Downloads
0
Stars
0
Active Installs
3
Versions
Install in OpenClaw
/install investment-advisory-workflow
Description
［何时使用］当用户需要投顾服务时；当用户说'最近 XX 怎么看'、'这个事件有什么用'、'帮我看看持仓'、'100 万怎么配置'、'大跌了怎么办'时触发。场景驱动的投顾全流程，融合四专家思维。
README (SKILL.md)
investment-advisory-workflow: 投资顾问工作流 🎯
📋 功能描述
帮助用户系统化执行投顾全流程。融合林奇 (洞察)/卡尼曼 (行为)/芒格 (逆向)/马利克 (系统) 四位专家思想，覆盖 5 个用户场景。
适用场景：
- 市场解读 / 事件分析 / 持仓诊断 / 资产配置 / 行为纠偏
边界条件：
- 不替代持牌投顾服务
- 输出为 Markdown 报告，需配合 data_layer / mcp-aktools / qieman-mcp 获取真实数据
- 场景识别依赖用户输入关键词与情绪表达
- KYC 前置：资产配置场景（场景 4）必须先收集用户年龄/风险偏好/金额/期限，不直接给配置方案
- 四专家标注：输出必须包含 [林奇视角]/[卡尼曼视角]/[芒格视角]/[马利克视角] 标注，确保思维融合
- 隐私保护：检测到敏感信息（身份证/银行卡）必须脱敏，不入库
🔄 5 个核心场景
| 场景 | 触发词 | 调用步骤 | 输出 |
|---|---|---|---|
| 市场解读 | "最近 XX 怎么看？" | market-scan → industry-rank → plain-explain → ljg-card | Markdown + PNG 卡片 |
| 事件分析 | "这个事件有什么用？" | market-scan → industry-rank → multi-view → plain-explain → decision-integrate | Markdown 影响分析 |
| 持仓诊断 | "帮我看看持仓" | data-query → holding-diagnoser → decision-checklist → fund-allocator → report-generator | Markdown 诊断报告 |
| 资产配置 | "100 万怎么配置？" | decision-checklist → fund-allocator → ljg-roundtable → IPS 模板 → report-generator | Markdown 配置方案 |
| 行为纠偏 | "大跌了怎么办？" | market-scan → companion-script → ljg-relationship → problem-mapper → plain-explain | Markdown 纠偏方案 |
详细四专家框架 → references/four-experts.md
六阶段能力库 → references/six-stages.md
共享 Skill 说明 → references/shared-skills.md
⚠️ 常见错误
错误 1：混淆投资工作流与投顾工作流
问题：
• 用户问"帮我看看持仓"，却调用投资工作流的 stock-research
• 输出偏重标的分析，忽略用户心理与行为纠偏
解决：
✓ 投顾工作流核心是"帮别人"，侧重 KYC + 行为干预 + 陪伴
✓ 投资工作流核心是"自己投"，侧重标的分析 + 决策验证
✓ 严格匹配场景定义
错误 2：忽略四专家视角融合
问题：
• 输出只有数据，没有行为纠偏或逆向思考
• 像数据报告，不像投顾建议
解决：
✓ 每个场景必须融合至少 2 个专家视角
✓ 标注 [林奇视角]/[卡尼曼视角]/[芒格视角]/[马利克视角]
✓ 输出包含"洞察 + 行为 + 逆向 + 系统"四维结构
错误 3：配置方案硬编码
问题：
• 直接给固定比例，不协商
• 忽略市场观点动态调整
解决：
✓ fund-allocator 必须输出基础配置 + 调整后配置
✓ 生成协商点（风险偏好 vs 配置比例）
✓ 标注调整理由与置信度
错误 4：情绪识别缺失（冷冰冰）
问题：
• 用户说"大跌了，我好慌"，AI 直接给数据报告
• 忽略情绪，导致建议"冷冰冰"，可能引发非理性操作
解决：
✓ 阶段 1 必须检测情绪词（"慌"、"割肉"、"好怕"、"大跌"）
✓ 若检测到情绪，优先调用 companion-script 安抚话术
✓ 原则：先处理情绪，再处理问题
错误 5：KYC 前置不足（无画像不配置）
问题：
• 用户问"100 万怎么配"，AI 直接给比例
• 未收集年龄/风险偏好/期限，配置方案不匹配
解决：
✓ 执行"无 KYC，不配置"原则
✓ 若信息不全，暂停配置流程，先引导用户完成 KYC 问卷
✓ 输出中必须包含"基于您的风险等级为 XX"的声明
错误 6：隐私保护缺失
问题：
• 用户输入身份证号/银行卡号，AI 原样输出或入库
• 数据安全风险
解决：
✓ 立即脱敏：输出时掩码处理（如"6222 **** **** 1234"）
✓ 安全提示：提醒用户"请勿在对话中发送完整身份证号/银行卡号"
✓ 不入库：敏感信息不写入知识库/日志
🧪 使用示例
输入：
最近消费怎么看？
预期输出：
- 识别场景：市场解读
- 调用：market-scan → industry-rank → plain-explain → ljg-card
- 输出：Markdown 解读 + PNG 卡片（含四专家视角标注）
输入：
大跌了，我好慌，要不要割肉？
预期输出：
- 识别场景：行为纠偏 + 情绪检测
- 调用：market-scan → companion-script（安抚）→ ljg-relationship（行为识别）→ problem-mapper（纠偏）
- 输出：Markdown 安抚话术 + 纠偏方案
输入：
100 万怎么配置？
预期输出：
- 识别场景：资产配置
- 调用：decision-checklist（KYC 问卷）→ fund-allocator → ljg-roundtable → IPS 模板 → report-generator
- 输出：Markdown 配置方案 + IPS（若 KYC 不全，先询问）
🔧 故障排查
| 问题 | 检查项 |
|---|---|
| 不触发 | description 是否包含触发词？用户输入是否匹配场景？ |
| 数据为空 | data_layer 是否安装？mcp-aktools/qieman-mcp 是否运行？ |
| 输出像投资报告 | 是否混淆投资工作流？检查场景定义与专家视角融合 |
| 配置无协商点 | fund-allocator 是否调用？是否生成协商点？ |
| 缺乏行为纠偏 | 是否调用 companion-script / ljg-relationship？ |
| 情绪未识别 | 是否检测情绪词？是否优先安抚？ |
| KYC 缺失 | 是否执行"无 KYC 不配置"？是否先询问画像？ |
| 隐私泄露 | 是否检测敏感信息？是否脱敏输出？ |
🔗 相关资源
- 四专家框架：
references/four-experts.md - 六阶段能力库：
references/six-stages.md - 共享 Skill 文档：
references/shared-skills.md - 报告模板：
templates/report-template.md - 标准参考：
docs/SKILL-STANDARD-v3.md
Usage Guidance
This skill appears coherent and instruction‑only, but before installing: (1) Audit the referenced shared skills/data sources (data_layer, mcp-aktools, qieman-mcp, any provider clients) to see what API keys, network endpoints, or logging they require; (2) Confirm platform policies for allowed-tools — if you don't want the agent to run shell commands or access the filesystem, restrict Exec/Bash/Read/Write permissions; (3) Verify how KYC/PII is collected, stored, and logged: test with synthetic/anonymized data and ensure desensitization is enforced and sensitive fields are not written to logs or external services; (4) Review compliance/disclaimer handling for regulatory requirements in your jurisdiction before using this for real client advice; (5) If you will connect live brokerage/data accounts, inspect those integrations' credential requirements and scope (least privilege) — the skill itself doesn’t request secrets but the shared modules it calls might. If you want higher assurance, request the concrete implementations of the shared skills and any network endpoints they contact and have them reviewed.
Capability Analysis
Type: OpenClaw Skill
Name: investment-advisory-workflow
Version: 1.3.0
The skill bundle defines a comprehensive investment advisory workflow focused on financial analysis and behavioral finance. It includes explicit instructions for privacy protection (masking sensitive data like ID and bank cards), KYC compliance, and structured financial reporting. While the skill requests high-privilege permissions such as 'Bash' and 'Exec', the logic in SKILL.md and shared-skills.md is strictly aligned with data retrieval via legitimate financial libraries (e.g., AKShare) and analysis; no evidence of malicious intent, data exfiltration, or prompt-injection attacks was found.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description describe a multi‑stage investment advisory workflow and the SKILL.md content, README, and templates consistently implement that workflow. The skill explicitly says it needs data from data_layer / mcp-aktools / qieman-mcp to produce real results; those dependencies are referenced in the manifest. The skill does not request unrelated credentials or system access in its own metadata, so required capabilities align with the stated purpose.
Instruction Scope
Runtime instructions cover scene routing, KYC gating, emotion detection, and composing multi‑view analyses and reports; these stay within an advisory workflow. The SKILL.md allows tools including Bash/Exec/Read/Write/WebSearch — reasonable for generating reports and images, but broader than strictly necessary. No instructions in SKILL.md tell the agent to read arbitrary system files or exfiltrate data; however the allowed-tools scope means the agent could be given broad file/exec rights at runtime, so operators should ensure the platform enforces tool permission limits.
Install Mechanism
This is an instruction‑only skill with no install spec and no code files, so nothing is written to disk by the skill itself. That is low risk and proportionate for a workflow description.
Credentials
The skill declares no required environment variables or primary credential, which is consistent with being an orchestration/instruction skill. It references external shared modules (data_layer, mcp-aktools, qieman-mcp, fund APIs) that may require credentials or network access — those are not part of this skill's metadata and should be audited separately. The SKILL.md also requires KYC data collection and explicit privacy/desensitization rules; ensure collecting/storing KYC is compliant and that underlying modules don't leak PII.
Persistence & Privilege
always:false and default model invocation settings are used. The skill does not request permanent presence nor does it declare any behavior that modifies other skills or system‑wide settings. Nothing in the package indicates it will persist credentials or alter agent configuration beyond normal operation.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install investment-advisory-workflow - After installation, invoke the skill by name or use
/investment-advisory-workflow - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.3.0
新增数据来源标注/四专家视角标注/置信度评估/操作建议具体化
v1.1.0
v1.1.0: 增加 KYC 前置流程，四专家视角标注，符合标准 3.0
v1.0.0
v1.0.0: 场景驱动版工作流，融合四专家思想，覆盖 5 个用户场景
Metadata
Frequently Asked Questions
What is Investment Advisory Workflow?
［何时使用］当用户需要投顾服务时；当用户说'最近 XX 怎么看'、'这个事件有什么用'、'帮我看看持仓'、'100 万怎么配置'、'大跌了怎么办'时触发。场景驱动的投顾全流程，融合四专家思维。 It is an AI Agent Skill for Claude Code / OpenClaw, with 99 downloads so far.
How do I install Investment Advisory Workflow?
Run "/install investment-advisory-workflow" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Investment Advisory Workflow free?
Yes, Investment Advisory Workflow is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Investment Advisory Workflow support?
Investment Advisory Workflow is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Investment Advisory Workflow?
It is built and maintained by lj22503 (@lj22503); the current version is v1.3.0.
More Skills