Workflows · VALUEFOCUS RESEARCH

Build an AI Agent for Portfolio Risk Analysis

Review concentration, correlated exposures and thesis risks without delegating portfolio decisions to an agent.

3 Tools covered

Tools covered

Investment Guru Skill

A Chinese-first analysis framework that applies methods associated with more than twenty investing masters.

Stock Research Framework

A nine-part stock-research checklist covering business, financials, competition, valuation, macro and sentiment.

Finance Skills

A mature collection of Agent Skills for financial analysis, data providers and read-only market research.

Recommended workflow

1
Verify the repository, license and latest release
2
Install in an isolated environment and keep credentials in environment variables
3
Enable read-only tools first and inspect every exposed capability
4
Run a sourced test query and confirm timestamps before relying on output

Examples and expected outcomes

Example 1

Create a repeatable research memo

Sourced evidence
Repeatable steps
Explicit uncertainty

Example 2

Compare companies with one checklist

Sourced evidence
Repeatable steps
Explicit uncertainty

Example 3

Build a sourced bull/bear review

Sourced evidence
Repeatable steps
Explicit uncertainty

Access requirements

Keys and runtime

No core key stated; network data sources are required

License and data rights

SkillNav lists it as unscanned; verify source files and data terms

Rate limits

Tencent/Yahoo-style endpoints may throttle or change

Connect to your Agent

Claude config

claude skill add --url https://github.com/openclaw/skills

# Project-scoped Agent Skill layout
.claude/skills/investmentguru-review/SKILL.md
.claude/skills/investmentguru-review/references/

# Restart Claude Code and confirm discovery.

Prompt example

“Use Investment Guru Skill to research AAPL. List sources and timestamps first, then separate facts, calculations, risks and missing data. Do not give personalized trading advice.”

When not to use it

SkillNav marks it unscanned
Displayed 4.5k popularity belongs to the source repository, not this individual skill
Master-style outputs are interpretations, not primary data

How to choose

Prefer primary sources for filings, maintained adapters for convenience, and narrow permissions for every agent tool. Treat model conclusions as analysis rather than financial advice.

Evidence to retain

Original source URL and retrieval timestamp

Exact tool inputs and normalized output

Missing-data and rate-limit messages

A clear separation between facts and interpretation