Investment Masters: Markdown Profiles of 15 Investors, Packaged as an Agent Skill
Distill investment wisdom from Buffett, Dalio, Soros, Marks & more — 15 masters' methodologies + 13F tracking, built for AI agents.
At a glance
- What is it?
- AlphaGBM's repository turns shareholder letters, memos and SEC filings into structured profile files that an AI coding agent can read on demand. The interesting part is the packaging, not the prose.
- Who is it for?
- Adopt this if you already run an agent with filesystem access and you want investor methodology as reference material inside that agent's context, rather than as a separate research workflow. Skip it if you expect the repository to fetch or parse EDGAR filings for you, because the README describes 13F tracking as CIK links, not as a downloader.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 112 days ago.
- What is it written in?
- GitHub does not report a main language for this repository.
Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
The problem is context, not investment data
Public investing material is not scarce. Shareholder letters, 13F filings, memos and books are all free. What is scarce is a format an AI agent can pull into a working session without a human pasting excerpts by hand. Investment Masters answers that narrow problem: it stores 15 fund managers' methodologies as structured markdown files under masters/, plus a SKILL.md that defines how an agent should use them. The README frames the goal as methodology extraction rather than impersonation, drawing a line between this and tools that generate quotes in a famous investor's voice. The audience is correspondingly narrow. It is for people who already work inside Claude Code or Cursor and want a named directory of investor profiles available to the model, and for human readers who prefer one file per manager over a book-length summary. It is not a screening tool, a portfolio tracker, or a data feed.
One markdown file per manager, plus a skill definition
The repository layout is the whole mechanism. There is a README.md, a SKILL.md that the README calls the AI agent skill definition, an MIT LICENSE, and a masters/ directory holding one file per investor: ackman.md, aqr.md, ark.md, and so on through the 15 names. The README's table maps each master to a style and a profile path, for example Bridgewater (Dalio) to risk parity and All-weather, Renaissance (Simons) to statistical arbitrage, and Raschke to short-term technical trading. The data flow is retrieval by filename. An agent that can list and read files resolves a question like "distill Buffett's methodology" to masters/buffett.md and reads it. Nothing in the supplied material describes an index, embeddings, or a query layer, so the practical unit of retrieval is the file itself. That is a deliberate simplicity, and it also sets the ceiling: an agent can only use what fits in context alongside the rest of the task.
Installing it means cloning into an agent's skill directory
There is no package manager step. The README gives two commands. For Claude Code: git clone https://github.com/AlphaGBM/investment-masters.git .claude/skills/investment-masters. For Cursor: git clone https://github.com/AlphaGBM/investment-masters.git .cursor/skills/investment-masters. The target path is the configuration. The repository lands where the tool looks for skills, and SKILL.md is what the agent reads to learn the skill's purpose. Updating is a git pull in that directory, since there is no release artefact and the repository's recent releases field is empty. The README does not document any environment variables, API keys, or config keys, which is consistent with a content-only repository: if a clone into the right folder is the entire setup, there is nothing to configure. The example prompts in the README ("Compare Dalio and Marks on risk management", "What did Bridgewater buy last quarter?") are the intended entry points, and they assume the agent has both the skill directory and, for the holdings question, the ability to follow links.
13F tracking is a table of CIK links, not a pipeline
This is where the README's framing runs ahead of its contents. The project advertises 13F tracking, and the section that delivers it is a table of seven institutions with CIK numbers and EDGAR browse-edgar URLs: Bridgewater at 0001350694, Berkshire at 0001067983, Appaloosa at 0001656456, Pershing Square at 0001336528, Soros Fund at 0001029160, Hillhouse at 0001510057, ARK Invest at 0001803918. Those links point at SEC EDGAR, which is free and public. What the repository supplies is the identifier mapping, not the retrieval. There is no described script that downloads a filing, parses the information table, or diffs one quarter against the next. So "Ackman's latest 13F changes" is a request the agent can only satisfy if it can browse the web and read a filing, and the quality of that answer depends on the agent, not on this repository. The README is also explicit that Renaissance Medallion and AQR's internal funds are proprietary and absent from 13F, which is honest and worth knowing before you wonder why two of the 15 names have no holdings row.
The five common principles are the most opinionated file
The README states that all 15 managers converge on five principles: systems over intuition, risk management over stock picking, a clear thesis with willingness to be wrong, cycle awareness, and long-term over short-term. Each is anchored to a named practice rather than a slogan. Dalio's All-weather and quadrants, Simons' quant models, AQR's factor frameworks, Marks' pendulum, Tepper's panic buying, Ackman's written exit criteria, Soros exiting when wrong. The fifth principle is the one that strains. The README itself flags the exception, noting that Renaissance profits from short-term statistical arbitrage, and the list includes Raschke's short-term technical style and Liang Wenfeng's quant approach at High-Flyer. A convergence claim that needs an explicit carve-out in its own text is doing rhetorical work rather than analytical work. Treat the five principles as a reading aid for the profiles, not as a finding the 15 managers would endorse.
What the material does not let you verify
The README claims every claim has a source citation and that profiles draw on primary sources: letters, 13F filings, books, memos, interviews. The supplied material shows the claim but not the citations, because no profile file's contents are included. The README also names specific source works per manager: 60 years of shareholder letters for Buffett, The Man Who Solved the Market for Simons, 200+ research papers for AQR, The Alchemy of Finance for Soros, The Most Important Thing plus memos for Marks, One Up on Wall Street for Lynch, Street Smarts and Market Wizards for Raschke, and Xueqiu essays for Duan Yongping. Those are reasonable pointers. Whether the prose in each file actually traces back to them is the one thing a reader cannot check from the README, and it is the first thing to check by opening two or three files. A profile that cites a book title in a header and then paraphrases from memory is not the same artefact as one that quotes with page or letter references.
The comparison that matters is against a wiki, not a chatbot
The README's own comparison table contrasts this project with AI roleplay tools and generic finance bots on source, depth, verifiability, actionability and 13F coverage. That table is the least useful part of the document, because the realistic alternative is not a chatbot. It is a personal notes vault or a wiki page per investor, which many people already maintain. The difference is packaging and invocation. A wiki page is something you open in a browser and read. A file at .claude/skills/investment-masters/masters/dalio.md is something an agent reads mid-task and reasons over, which is the actual reason to keep the content in this form rather than in Notion. The cost is that the content is only as good as the markdown, and it is versioned in git like code, so edits and disagreements are visible in history. If you would not maintain a folder of investor notes, this repository does not change that.
Licence, maintenance and what to check before trusting a profile
The repository is MIT licensed, which covers the repository's own text and structure. It does not cover the underlying sources the profiles summarise. Shareholder letters, books, memos and SEC filings carry their own rights, and a profile that paraphrases or quotes them inherits whatever constraints apply to that source. That is a question for a lawyer, not for this article, but it is worth noting that a permissive licence on the wrapper says nothing about the contents. On maintenance, the last push is 2026-05-27 and there are no retrieved releases, so the update path is git pull against main. The 13F links age every quarter, and the profiles themselves do not expire in the same way, which means the repository has two different clocks running. The concrete first step is to clone it into the skill directory your agent already uses, open masters/buffett.md alongside masters/duan_yongping.md, and judge whether the sourcing is consistent across a famous name and a less documented one. That comparison tells you more than any summary of the README can.
Editorial conclusion
Adopt this if you already run an agent with filesystem access and you want investor methodology as reference material inside that agent's context, rather than as a separate research workflow. Skip it if you expect the repository to fetch or parse EDGAR filings for you, because the README describes 13F tracking as CIK links, not as a downloader. Before relying on any profile, open masters/buffett.md and one of the less famous entries, such as masters/raschke.md, and check whether the sourcing style holds up across the set rather than only for the names you already know well.
Community notes