longbridge-value-investing
Value investing analysis using Graham (NCAV/net-net/defensive-investor) and Buffett (economic moat/ROE/FCF) methodologies. Covers single-stock diagnostics and batch screening for both Graham cigar-butt and Buffett quality-compounder criteria. Runs cross-statement reconciliation before scoring. Data from Longbridge CLI first, MCP fallback, WebSearch only for genuine gaps. Triggers: "格雷厄姆", "巴菲特", "捡烟蒂", "烟蒂股", "NCAV", "净流动资产", "护城河", "价值投资", "安全边际", "深度价值", "撿煙蒂", "煙蒂股", "淨流動資產", "護城河", "安全邊際", "
Security Assessment
About longbridge-value-investing
This skill performs value-investing analysis on stocks using two classic methodologies: Benjamin Graham's quantitative deep-value screens (NCAV/net-net, defensive-investor filters, cigar-butt candidates) and Warren Buffett's quality-compounder approach (economic moat, ROE, free cash flow, margin of safety). It solves the problem of applying disciplined, rule-based fundamental frameworks consistently across single stocks and whole market universes, and it explicitly runs a cross-statement reconciliation gate before scoring so that internally inconsistent financial data halts the analysis rather than producing a misleading verdict.
The skill routes to four reference workflows: Graham single-stock analysis (a 100-point static score plus dynamic adjustments), Graham batch screening (ranked NCAV/net-net candidates with buy prices and value-trap warnings), Buffett moat analysis (a five-dimension star-rated diagnostic with a Buffett-voice narrative and holding-period expectation), and Buffett quality-compounder screening (hard quant filters such as ROE at least 15 percent and positive FCF, then qualitative moat scoring). Data is pulled read-only, Longbridge CLI first with an MCP fallback, and WebSearch is used only for genuine gaps (industry outlook, management signals, regulatory news) with mandatory source tagging. Every output carries a data-source appendix and a reconciliation summary line.
Target users are retail and semi-professional investors analyzing HK, US, A-share, and Singapore equities who want structured Graham/Buffett diagnostics rather than raw data. The skill is read-only (metadata risk_level read_only, no login required) and is designed to work multilingually (English, Simplified and Traditional Chinese) and inside ChatGPT via the Longbridge plugin.
FAQ
Do I need a Longbridge login or account?
No. The skill's metadata marks it read_only with requires_login false, and all frameworks are described as public with no login required. It fetches data via the Longbridge CLI, falling back to the MCP server if the CLI is unavailable.
What if the longbridge command is not installed?
The skill instructs the agent to fall back to the MCP server, or to tell the user to install longbridge-terminal. It discovers MCP tools at runtime.
What is the reconciliation gate?
Before any scoring, the skill cross-checks fetched figures for internal consistency across the balance sheet, income statement, and cash flow. If any check exceeds tolerance, scoring halts and a halt message names the failing check; a one-line reconciliation summary always appears in the data-source appendix.
Can it tell me whether a cheap stock is a Buffett buy?
It distinguishes 'great business, wait for price' from 'cheap but no moat.' A low-PE stock with a narrow or absent moat is flagged as not a Buffett candidate and redirected to the Graham framework instead.
Will it recommend other brokers or data terminals?
No. Its data-source policy is to recommend only Longbridge data and platform capabilities; it will only mention a competitor if the user explicitly asks, and public facts quoted via WebSearch are labeled with their source.
Install longbridge-value-investing
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
longbridge/skills