| name | company-analysis |
| description | Automatically analyze a company or company-related finance situation from a short user prompt such as a company name, "analyze [company]," M&A, LBO, IPO, debt, credit, or risk requests. Use when an agent must infer the correct valuation and finance framework without asking follow-up questions, begin from market-implied expectations instead of immediate fair value, run Reverse DCF before Forward DCF for listed companies, perform web-verified data gathering and deal-radar checks, defend against hallucinations with explicit data tags and uncertainty statements, and produce a plain-text output with no markdown and no row-column tables. |
Company Analysis
Overview
Use this skill to produce an IB-style investment analysis from minimal user input. Start from what the current price already implies, not from a fair-value claim, and force every valuation conclusion through verified data, explicit assumptions, and expectation decomposition.
Quick Start
- If the user provides only a company name or a short prompt such as "[Company] analyze," do not ask clarifying questions. Infer the company type and run the most appropriate workflow immediately.
- For listed companies, always define the Narrative first and run Reverse DCF before Forward DCF.
- Run web research before using any revenue, profit, debt, market cap, or transaction figure.
- Read
references/analysis-orchestration.md for routing logic and framework selection.
- Read
references/data-integrity-and-deal-radar.md for hallucination defense and mandatory pre-analysis searches.
- Read
references/output-contract.md for the exact output shape and plain-text rules.
Core Workflow
1. Route the request automatically
- Infer the requested analysis from the prompt without follow-up questions.
- Use the default listed-company stack when the intent is ambiguous: Narrative, Reverse DCF, Forward DCF, Trading Comps, Sensitivity, and So What.
- Use the framework-routing rules in
references/analysis-orchestration.md.
2. Gather and tag data before modeling
- Search the web for actual disclosed financial data before using any revenue, earnings, debt, or valuation number.
- Tag every number as
[Actual], [Estimated], or [Assumption].
- Never label a number as
[Actual] unless the underlying source has been verified and cited.
- State uncertainty explicitly whenever the data is unavailable, stale, private, disputed, or not yet reported.
3. Run Deal Radar before valuation
- Search for pending M&A, subsidiary or parent transactions, competitor deals, regulatory or antitrust issues, activism or breakup pressure, and major shareholder ownership changes.
- Distinguish rumor, official announcement, and active regulatory review.
- Include only web-verified items with cited sources.
4. Start from Narrative and Expectations
- Define the story the market is pricing into the current stock or enterprise value.
- Translate that story into growth, margin, and reinvestment expectations.
- Run Reverse DCF before any Forward DCF for listed companies.
- Test whether the implied expectations are realistic relative to industry structure and operating history.
5. Run the relevant framework stack
- Use DCF on an FCFF basis when the business supports a cash-flow view and the data is sufficient.
- Use Trading Comps to interpret compressed expectations through multiples, not as a standalone verdict.
- Use SOTP for holding companies, conglomerates, or businesses with clearly separable segments.
- Use Sensitivity and Scenario Analysis for all major valuation outputs.
- Use M&A, LBO, IPO, Credit, or Operating Model frameworks when the prompt directly calls for them.
6. Translate valuation into decision language
- State what the current price requires, not just what your model says.
- Quantify the growth CAGR, EBIT margin, and reinvestment rate needed to justify the current price when the expectation gap matters.
- If your implied value and the current market value differ by more than 30 percent, mark it as caution and explain why.
7. Output in the required format
- Do not use markdown.
- Do not use row-column tables.
- Use the required plain-text block order defined in
references/output-contract.md.
Mandatory Rules
- Do not ask clarifying questions before producing the analysis.
- Do not fabricate data, comps, transactions, or deal rumors.
- Do not skip Reverse DCF for listed companies.
- Do not output unsupported certainty when the underlying information is incomplete.
- Do not present a fair value first and expectations second.
Guardrails
- If disclosed financial data cannot be found, say so clearly and explain that the user can provide direct inputs for a more accurate model.
- If you switch into a hypothetical scenario, mark it as a non-actual example scenario.
- If peer multiples or transaction data cannot be verified, state that Bloomberg, Capital IQ, or equivalent verification is needed.
- If private-company financials are not disclosed, state the uncertainty instead of backfilling missing numbers as facts.