Create detailed business overviews for investment research — deep-dive through business model, management, financials, valuation, risks and tracking. Use this skill only when the user explicitly asks to do a business overview, deep dive, or formal company write-up.
Create detailed business overviews for investment research — deep-dive through business model, management, financials, valuation, risks and tracking. Use this skill only when the user explicitly asks to do a business overview, deep dive, or formal company write-up.
allowed-tools
yfinance_*, ibkr_*, fred_*, qmd_*, Read, Bash
Business Overview
CRITICAL RULE (Never Breach)
Never save a business overview until EVERY heading has been researched and discussed back-and-forth with the user. At each heading: surface what you see, suggest angles, then stop and ask — do not proceed until the user confirms. If the user tries to rush, push back.
Adapt the Approach
The business type determines the starting point — deep value starts with balance sheet and downside floor, high-ROIC stable growth starts with FCF and returns. Stable cash flows are always relevant regardless of type. Let the back-and-forth discussion surface what matters; do not impose a fixed checklist of metrics.
Workflow
Before diving into each heading, briefly state what you're looking for and why given the business type. Only tick the heading when the user signals they're satisfied — never assume a heading is done without explicit confirmation.
The Business — Moat, revenue model, economic environment, capital cycle.
The Management — Track record, insider ownership, skin in the game.
The Numbers — Key metrics tailored to the business type.
Valuation — Conservative assumptions, upside, expected return. If it cannot be valued on the back of a napkin, move on.
Risks — Key threats, probability, magnitude.
Tracking — What confirms the thesis? What invalidates it? Concrete thresholds and kill criteria.
Conclusion — 2-3 sentences synthesising the discussion.
Sources — Direct links to the actual documents used, not landing pages.
Save — Only after every heading above is confirmed.
Valuation Models
The user works with two approaches depending on business type:
DCF and reverse DCF — for high-ROIC stable growth businesses. Use reverse DCF to let the market tell you what assumptions are priced in, then debate whether they're too optimistic or pessimistic.
Balance sheet deep-dive — for deep value. Focus on net cash, receivable recovery, inventory liquidation, PP&E recovery, goodwill write-offs, off-balance-sheet items (operating leases, pensions, contingencies), and the user's own recovery estimates per asset class. Determine the hard downside floor before looking at earnings power.
Output Template
Use this structure when writing the final note:
Title format:# Company: A Descriptive Label (e.g., "Emeco: A Deleveraged Turnaround")
## The Business
What the company does, its moat, and revenue model. Assess capital cycle — is the industry starved or over-supplied?
## The Management
Track record, insider ownership, and incentives. Are they owner-operators with skin in the game?
## The Numbers
Key metrics, tailored to the business: revenue growth, margins, ROIC, debt, goodwill, etc. Focus on cash flows, returns on capital, and balance sheet strength.
## Valuation
Conservative assumptions (growth rate, exit multiple, margin of safety). Quantify upside and expected return. If it cannot be valued on the back of a napkin, move on.
## Risks
Key threats to the thesis: competition, cyclicality, balance sheet, management execution. Consider probability and magnitude.
## Tracking
Metrics to monitor. What confirms the thesis? What invalidates it? Use concrete thresholds.
## Conclusion
2–3 sentences: why attractive at today's price, long-term outlook.
## Sources
Direct links to sources used in the analysis — annual reports, ASX filings, industry data, company websites, news articles. No landing pages or generic investor relations URLs.
Once the final note has been written, use surgical edits only.
Gotchas
Back-and-forth is the engine. One message per heading is never enough. Probe, challenge, adapt — do not arrive with a fixed question list.
Source everything. Cite the actual document ("Annual Report FY2025," "Q2 earnings call transcript"), not the tool used to fetch it. If data is missing (segment detail, off-balance-sheet items, insider specifics), ask the user if they have the source. Do not make assumptions to fill gaps.
Do not fill the template during discussion. Only populate at the end after all headings are exhausted.
Third person only. Analysis, not personal opinion.
Information-dense, not terse. Pack analytical weight into every line, but maintain narrative flow. Data should do the heavy lifting — commentary is seasoning, not filler.