Guide · Agent skill
How to use the Private-Company Research Framework agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The private-company-research skill is a set of instructions that teaches an AI agent how to research an unlisted company, such as a pre-IPO firm, and estimate what its business is worth. It is built for investors and analysts preparing for a funding round or secondary purchase, and for students who want to see how private-company valuation is approached. By the end of this guide you will know what the skill does, how to install it, how to ask for it in plain language, and where its limits are.
Private-Company Research FrameworkHKUDS/Vibe-Trading · MIT Download the skill Quick start
- 1Download the ZIP. Get the private-company-research ZIP from this page and unzip it to reveal SKILL.md.
- 2Choose an install location. Place the folder in ~/.claude/skills/ for personal use, or in a project's .claude/skills/ folder for one project.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the ZIP.
- 4Confirm your agent has search. Make sure the agent has run_swarm or web_search and internet search access.
- 5Name the unlisted company. Ask for research on a specific private company and state the decision you are making.
- 6Review confidence labels. Check each high, medium and low label and the sources behind it before relying on any figure.
- 7Verify the gaps independently. Use the information-gap map to decide what to check with primary sources before acting.
What an agent skill is and what this one makes the agent do
An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when a task matches the skill's description. This skill is designed for one job: judging an unlisted company where public market prices do not exist.
The skill splits the work into six analyst lenses: business model and users, financial reconstruction, competitive landscape, risk and governance, technology and IP, and alternative signals such as hiring, patents, litigation, app activity or supply chain. The lenses are meant to run in parallel through run_swarm, one worker per lens, or one after another using web_search. The results are then cross-validated for consistency before any verdict is given.
- Every data point carries a confidence label: high, medium or low.
- Verifiable facts are kept separate from the agent's own inference.
- "I don't know" is treated as a valid answer rather than a gap to fill with guesses.
- Intended outputs: a fair-value range, an exit-path analysis, and an information-gap map.
Install the skill from a ZIP file
Download the ZIP from this page and unzip it. The package contains a single file, SKILL.md. You can install it in one of three places, depending on how you use your agent.
For a personal setup in Claude Code, place the unzipped skill folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder. In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/ (available across projects).
- Project install: .claude/skills/ inside the project folder (available only there).
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP.
Check what the skill needs before you run it
The skill needs an agent with sub-agent support (run_swarm) to run the lenses in parallel, or web_search to run them one by one. It also needs internet search access, because the research depends on gathering public sources.
The skill does not run scripts. It does use the network, since it searches for sources on the web.
- Agent with run_swarm for parallel lenses, or web_search for sequential research.
- Internet search access to collect sources.
- Runs scripts: no. Uses network: yes.
Trigger the skill with natural requests
You do not need a special command. Describe an unlisted company and what you need to know about its value. The skill's description covers requests about any unlisted company where you need to judge what the business is worth.
Name the company and say what decision you are making. Asking for the information gaps, not just a number, is useful because the skill is built to show what it could not confirm.
- "Research Stripe's business and tell me a fair-value range, with confidence labels on every number."
- "I'm considering a secondary purchase in a pre-IPO company. Run the private-company research framework and map the information gaps."
- "Assess ByteDance's competitive position and possible exit paths, and flag any single-source claims."
- "Using the private-company framework, what would have to be true for this startup to succeed, and what is the key validation point at its stage?"
Read the output with the confidence labels in mind
Expect each figure to be tagged high, medium or low confidence, with its source and how it was derived. Facts and inferences should appear in separate places. Where evidence is thin, the report should say so rather than present a precise number.
The skill also warns that online information leans toward company-friendly news. Treat any high-confidence label as the agent's judgement about the sources it found, not as audited financial fact.
- Check the source behind each high-confidence number.
- Treat single-source data as weak even if it is labelled well.
- Use the information-gap map to decide what to verify before committing money.
What people use it for
- Preparing a view on a pre-IPO company before a funding round.
- Assessing the risks and governance of a firm before buying secondary shares.
- Learning how private-company valuation is structured, lens by lens.
- Building a checklist of unknowns to bring to a due-diligence call.
- Comparing the competitive position of several unlisted companies in the same sector.
Frequently asked questions
Is the private-company-research skill free?
The source material for this skill is released under the MIT licence, which is stated on this page. Check the licence file in the repository for the full terms before redistributing it.
What do I need to run the private-company-research skill?
You need an agent with sub-agent support (run_swarm) or web_search, plus internet search access. The skill does not run scripts but does use the network to find sources.
How do I install an agent skill from a ZIP file?
Unzip the download and place the folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder. In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
Can this skill value a listed company like Apple?
No. The skill is scoped to unlisted firms. For listed companies, standard public filings and market data are a better basis.
Why does the skill say 'I don't know' so often?
By design, it prefers leaving blanks over filling tables with speculation. Private companies publish little, so the skill marks gaps instead of disguising guesses as sourced analysis.
Is the output enough to make an investment decision?
No. The skill relies on web search and public reporting, and its own text warns that online information is biased toward company-friendly news. Verify key figures independently before committing money.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are an investor, analyst or student who needs a disciplined, confidence-labelled look at an unlisted company before a funding round or secondary purchase. Pick something else if you need a decision-ready answer on a listed company, since public filings and market data are a better basis, or if you expect a verified valuation without doing your own checks.
Download the skill