20 ChatGPT-5.5 Prompts for Financial Analysts: DCF Models, Earnings Analysis, Risk Assessment, and Investment Research

Prompt 1 — Comprehensive 10-Year DCF Model from Historical Financials to Fair Value

Production-ready Prompt Text:

Act as a professional equity analyst. Using the provided historical financials and market inputs, construct a 10-year discounted cash flow (DCF) model. Return output in VALID JSON with keys: assumptions, projections (yearly), wacc (detailed breakdown), terminal_values (Gordon Growth and Exit Multiple), valuation (enterprise value, equity value, per-share fair value), and sensitivity_table. Include calculation formulas and rounding conventions. Avoid extraneous commentary.

Context Variables

• company_name: “Acme Co.”
• fiscal_year_end: “2025-12-31”
• historical_revenue: {2021: 2,100,000; 2022: 2,450,000; 2023: 2,900,000; 2024: 3,300,000} USD
• historical_operating_income, capex_schedule, depreciation_schedule
• working_capital_rules: days receivable, payable; tax_rate (e.g., 21%)
• risk_free_rate (4.10%), market_premium (6.0%), beta (1.05), pre_tax_cost_of_debt (6.5%), target_debt_to_total_capital (25%)
• terminal_Gordon_Growth_rate (2.5%), exit_multiple (e.g., 10.5x NTM EBITDA)
• shares_outstanding (excluding dilutive instruments), net_debt (cash minus debt)

Expected Output Format (JSON Schema)

{
  "assumptions": {
    "discount_rate_description": "WACC calculated using CAPM and cost of debt",
    "tax_rate": 0.21,
    "growth_assumptions": {"year1": 0.08, "year2": 0.06, ...}
  },
  "projections": [
    {"year": 2026, "revenue": 3300000, "unlevered_fcf": 265000, ...},
    ...
  ],
  "wacc": {
    "cost_of_equity": 0.085,
    "cost_of_debt_after_tax": 0.0525,
    "wacc": 0.075
  },
  "terminal_values": {
    "gordon_growth": 123456789,
    "exit_multiple": 130000000
  },
  "valuation": {
    "enterprise_value": 250000000,
    "equity_value": 200000000,
    "per_share": 24.50
  },
  "sensitivity_table": {
    "g_growth": {"1.5%": ..., "2.5%": ...},
    "wacc": {"7.0%": ...}
  }
}

Usage Notes

• Maintain temperature at 0.0 for deterministic, consistent numerical results
• Validate terminal EV/EBITDA multiples—values above 30x merit review
• For complex outputs, paginate requests to stay within token constraints

Example Input Snippet

{
  "company_name": "Acme Co.",
  "fiscal_year_end": "2025-12-31",
  "historical_revenue": {"2021": 2100000, "2022": 2450000, "2023": 2900000, "2024": 3300000},
  "tax_rate": 0.21,
  "shares_outstanding": 8200000,
  "net_debt": 50000000,
  "risk_free_rate": 0.041,
  "market_premium": 0.06,
  "beta": 1.05
}

Example Output (Abridged)

{
  "assumptions": {"tax_rate": 0.21, "terminal_growth": 0.025, "wacc": 0.075},
  "projections": [{"year": 2026, "unlevered_fcf": 265000}, {"year": 2027, "unlevered_fcf": 285500}, ...],
  "wacc": {"cost_of_equity": 0.085, "cost_of_debt_after_tax": 0.0525, "wacc": 0.075},
  "terminal_values": {"gordon_growth": 125000000, "exit_multiple": 138000000},
  "valuation": {"enterprise_value": 260000000, "equity_value": 210000000, "per_share": 25.61}
}

Prompt 2 — Dynamic WACC Calculator with Sensitivity Analysis

Production-ready Prompt Text:

Calculate a detailed weighted average cost of capital (WACC) from inputs provided. Compute cost of equity using CAPM (cost_of_equity = risk_free_rate + beta * market_premium), pre-tax and after-tax cost of debt, and capital structure weights. Generate a sensitivity matrix varying beta (0.8 to 1.6, step 0.1) and market premium (4.0% to 8.0%, step 0.5%). Output a CSV-compatible table with columns: beta, market_premium, wacc.

Context Variables

• risk_free_rate (e.g., 4.10%)
• base_beta (1.05)
• market_premium_base (6.0%)
• pre_tax_cost_of_debt (6.5%)
• tax_rate (21%)
• debt_weight (25%), equity_weight (75%)

Expected Output Snapshot

{
  "base_wacc": 0.075,
  "components": {
    "cost_of_equity": 0.085,
    "cost_of_debt_pre_tax": 0.065,
    "cost_of_debt_after_tax": 0.05135
  },
  "sensitivity_grid": [
    {"beta": 0.8, "market_premium": 0.04, "wacc": 0.0623},
    {"beta": 0.9, "market_premium": 0.045, "wacc": 0.0644},
    ...
  ]
}

Best Practice Tips

• Regularly update market premium and risk-free rate using authoritative sources
• Recalculate WACC when adjusting capital structure or tax assumptions

Prompt 3 — Terminal Value Methodology Comparison and Sensitivity Analysis

Production-ready Prompt Text:

Calculate terminal value using: (a) Gordon Growth Model with specified growth rate and WACC, and (b) Exit Multiple methodology leveraging peer EV/EBITDA multiples. Normalize peer multiples by computing median and quartiles (25th, 75th percentiles). Provide a comparison table with implied per-share values under three valuation scenarios: conservative (25th percentile), base (median), and aggressive (75th percentile). Return JSON including peer multiples used.

Context Variables

• last_projected_ebitda (e.g., 120,000,000 for 2029)
• peer_ev_ebitda_list (e.g., [8.5, 9.8, 11.2, 14.0, 10.3])
• wacc, terminal_growth_rate
• shares_outstanding, net_debt

Expected Output Layout

{
  "peer_stats": {"25th": 9.3, "median": 10.3, "75th": 11.6},
  "terminal_values": {
    "conservative": 950000000,
    "base": 1236000000,
    "aggressive": 1392000000
  },
  "per_share": {
    "conservative": 12.34,
    "base": 16.04,
    "aggressive": 18.08
  }
}

Usage Notes

• Flag terminal multiples that imply unrealistic long-term growth beyond economic fundamentals
• When peer sample size is small (<5), prompt model to augment peer set or warn on confidence

Prompt 4 — Free Cash Flow Bridge and Three-Statement Reconciliation

Production-ready Prompt Text:

Given income statement, balance sheet, and cash flow statement data (annual or quarterly), reconcile net income to unlevered free cash flow (FCF) per period. Provide a stepwise bridge listing net income, depreciation & amortization, capex, changes in working capital, tax effects, and resulting unlevered FCF. Output CSV-compatible table and JSON array with keys: period, net_income, d_and_a, capex, wc_change, unlevered_fcf.

Context Variables

• periods: [“2023-Q4”, “2024-Q1”, “2024-Q2”, …]
• net_income, depreciation_and_amortization, capex, changes_in_working_capital, interest_expense (for tax adjustment)

Expected Output Format

[
  {"period": "2024-Q1", "net_income": 34000000, "d&a": 6500000, "capex": 12000000, "wc_change": -1500000, "unlevered_fcf": 37000000},
  ...
]

Usage Notes

• Explicitly flag missing or negative working capital line items; fail the prompt with error messaging if key data is omitted

Prompt 5 — Batch Scenario and Sensitivity Generator for DCF Models

Production-ready Prompt Text:

Generate three detailed DCF valuation scenarios: Base, Bear, and Bull, based on supplied assumptions. Include for each scenario a 10-year free cash flow projection, WACC, terminal value (Gordon Growth), enterprise and equity valuations, per-share fair value, and a concise 2-3 sentence narrative summarizing key drivers. Output structured JSON keyed by scenario name, including sensitivity matrices for WACC versus terminal growth rate.

Context Variables

• base_assumptions (growth rates, margins)
• downside_adjustments (e.g., margin minus 200bps)
• upside_adjustments (e.g., margin plus 150bps)
• scenario_probabilities (optional; e.g., base 50%, bear 30%, bull 20%)

Expected Output Structure

{
  "Base": {"fcf": [...], "wacc": 0.075, "terminal": 120000000, "equity_value": 200000000, "per_share": 24.4, "narrative": "Base case forecasts margin expansion..."},
  "Bear": {...},
  "Bull": {...}
}

Usage Notes

• Leverage scenario probabilities to compute weighted average equity valuations where applicable
• Ensure scenario narratives clearly link quantitative drivers such as revenue growth, margin shifts, and capex changes without vague language

Prompt 6 — Earnings Call Summary and KPI Extraction

Production-ready Prompt Text:

Extract and summarize the top 8 actionable items from the provided earnings call transcript or bullet points into a JSON structure: {items: [{title, type (guidance/product/operations/finance), quantitative_impact_estimate, confidence_level (low/medium/high), citation (timestamp or paragraph ID)}]}. Also extract numeric updated guidance ranges. Output JSON only.

Context Variables

• transcript_text (UTF-8, abide by token limits)
• recent_guidance (previous guidance figures for comparative analysis)

Expected Output Format

{
  "items": [
    {"title": "Revenue guidance raised", "type": "guidance", "quantitative_impact_estimate": "+3.5% FY revenue", "confidence": "high", "citation": "Q&A 12:34"},
    ...
  ],
  "updated_guidance": {"FY2026_revenue": [320000000, 330000000]}
}

Implementation Notes

• For transcripts exceeding token limits, segment by themes (e.g., prepared remarks, Q&A), summarize each, then synthesize in aggregate
• Attach precise timestamps or paragraph IDs to enable traceability of quantitative figures

Prompt 7 — Reconciliation of Consensus Estimates vs. Reported Earnings

Production-ready Prompt Text:

Compare the quarterly reported financials to provided consensus datasets. Produce a reconciliation table with columns: line_item, consensus_value, reported_value, absolute variance, percentage variance, and explanatory commentary highlighting drivers of variance. Include non-GAAP adjustments for EPS and EBITDA, quantifying their impact on headline metrics. Output JSON and CSV string.

Context Variables

• consensus (line items like revenue, EPS, EBITDA)
• reported_results (matching keys)

Expected Output Example

{
  "comparison": [
    {"line_item": "revenue", "consensus": 320000000, "reported": 327500000, "var_abs": 7500000, "var_pct": 2.34, "commentary": "Stronger pricing power in North America"}
  ],
  "csv": "line_item,consensus,reported,var_abs,var_pct,commentary\n..."
}

Validation Advice

• Flag variances exceeding two standard deviations based on the trailing eight quarters, highlighting potential anomalies

Prompt 8 — Non-GAAP to GAAP Reconciliation and Metric Normalization

Production-ready Prompt Text:

Using reported non-GAAP metrics and disclosed adjustments, generate a reconciliation table for GAAP metrics (EPS, EBITDA). Normalize by removing one-off items. Output JSON with fields: gaap_metric, non_gaap_metric, adjustments (list), and normalized_metric.

Context Variables

• non_gaap_adjustments (named with amounts and detailed descriptions)
• reported_gaap_metrics

Expected Output Sample

{
  "reconciliation": [
    {"metric": "net_income", "gaap": 35000000, "non_gaap": 37000000, "adjustments": [{"name": "restructuring", "amount": 2000000}]}
  ],
  "normalized": {"adjusted_eps": 1.24, "adjusted_ebitda_margin": 0.185}
}

Usage Insights

• Flag adjustments exceeding 1% of revenue or recurring across multiple periods—potential signs of structural shifts

Prompt 9 — Quarter-over-Quarter KPI Trend Analysis with Statistical Signaling

Production-ready Prompt Text:

Analyze quarterly KPIs spanning 12 quarters. For each KPI (e.g., revenue, gross margin, accounts receivable days, churn rate, ARPU), calculate QoQ and YoY percentage changes, 4-quarter rolling averages, and Z-scores relative to 12-quarter means. Output a JSON array with KPI analyses, concise directional signals (improving/declining/stable), and recommended next actions.

Context Variables

• kpi_time_series (e.g., {“revenue”: […], “gross_margin”: […], “churn”: […]})
• reporting_calendar (quarterly timestamp mapping)

Expected Output Example

[
  {"kpi": "revenue", "qoq_change": "1.8%", "yoy_change": "6.5%", "rolling_avg": "3.1%", "z_score": 1.4, "signal": "improving", "action": "Monitor product pricing trends"}
]

Implementation Tips

• Utilize rolling averages to smooth seasonality and prompt model to report statistical sample sizes

Prompt 10 — Professional Earnings Note Draft for Analyst Distribution

Production-ready Prompt Text:

Draft a 300–450 word earnings note for client dissemination, utilizing provided metrics and reconciliations. Organize the note with: key facts (bullet points), financial highlights (tabular bullets), investment implications (three bullets covering catalysts, risks, valuation), and analyst stance (Buy/Hold/Sell with price target and rationale). Include a brief methodology statement referencing model version and data sources. Output JSON keys: headline, key_facts, financial_highlights, implications, stance, methodology.

Context Variables

• headline_metrics (EPS, revenue, guidance updates, percentage variances)
• analyst_prior_rating_and_target

Expected Output Sample

{
  "headline": "Acme Co. beats Q2 estimates; raises guidance",
  "key_facts": ["Revenue $327.5m (+2.3% q/q)", "EPS $0.48 vs $0.42 est"],
  "financial_highlights": [{"line": "Revenue", "value": "327.5"}],
  "implications": [{"catalyst": "New product ramping in Q4", "risk": "Supply chain constraints affecting margins"}],
  "stance": {"rating": "Buy", "new_target": 28.00},
  "methodology": "DCF v3.1; data sourced from company release and consensus via Refinitiv; model date 2026-07-12"
}

Practical Advice

• Specify the tone (neutral, concise, or promotional) and target word count for consistent narrative style

Prompt 11 — Factor Attribution for Portfolio Excess Returns

Production-ready Prompt Text:

Using monthly returns over 60 months for a portfolio and its benchmark, along with factor exposures (market beta, size, value, momentum), perform a factor attribution decomposition. Return JSON with total_return, benchmark_return, active_return, allocation_effect, selection_effect, interaction_effect, plus factor_contributions (factor:contribution list). Summarize top drivers in a brief, three-line professional commentary.

Context Variables

• portfolio_returns (60-element monthly array)
• benchmark_returns (60-element monthly array)
• factor_returns (matrix: factors × months)
• current_factor_exposures (vector)

Expected Output Sample

{
  "total_return": 0.182,
  "benchmark_return": 0.145,
  "active_return": 0.037,
  "factor_contributions": [{"market": 0.025}, {"size": 0.004}, {"value": -0.002}]
}

Best Practice Tips

• Specify attribution methodology used (e.g., Brinson-Fachler) in output
• Validate total factor contributions sum to active return within a small rounding tolerance (e.g., 1e-4)

Prompt 12 — Portfolio VaR & CVaR Estimation Using Historical and Parametric Methods

Production-ready Prompt Text:

Estimate 1-day and 10-day Value at Risk (VaR) and Conditional VaR (CVaR) for the portfolio using: (1) historical simulation over 500 trading days and (2) parametric variance-covariance approach assuming normal and t-distributions (df=5). Output JSON with VaR and CVaR at 95% and 99% confidence levels, assumptions, and include a concise Python snippet illustrating calculations.

Context Variables

• positions: list of {ticker, weight, price_series}
• return_series: daily returns aligned by ticker
• holding_period_days: 1 or 10

Output Example

{
  "historical_VaR": {"1d": {"95%": 0.024, "99%": 0.042}, "10d": {...}},
  "parametric_VaR_normal": {...},
  "parametric_VaR_tdist": {...},
  "python_snippet": "import numpy as np\n..."
}

Usage Recommendations

• Check for return series stationarity and flag structural breaks (e.g., 2020 pandemic) that may bias historical simulation outputs

Prompt 13 — Custom Stress Test Scenario Generation for Portfolios

Production-ready Prompt Text:

Generate five bespoke stress-test scenarios tailored to the client portfolio considering sector weights, geographic exposures, leverage, and liquidity constraints. For each scenario provide: scenario_name, macroeconomic shock description (including GDP, interest rates, FX, commodities), expected P&L impact (% portfolio NAV), rationale, and recommended hedging strategies. Output JSON reflecting historical extremes and near-term credible risks as of July 2026.

Context Variables

• sector_weights, geographic_exposure, leverage
• holding_period_days, liquidity_constraints

Expected Output Sample

{
  "scenarios": [
    {
      "name": "Rapid Rate Spike",
      "macro_shock": "10-year rates +150 bps",
      "pnl_impact": "-4.2%",
      "hedge": "Increase duration hedges via interest rate swaps"
    }
  ]
}

Best Practice Guidelines

• Include actionability scores (1-5) and estimated hedging costs per scenario

Prompt 14 — Portfolio Optimization Using Mean-Variance with Constraints

Production-ready Prompt Text:

Optimize portfolio allocations using expected returns, covariance matrix, and portfolio constraints. Enforce long-only weights, full investment, asset-level bounds, sector caps, turnover limits, and a target volatility ceiling supplied by the user. Return valid JSON containing optimized weights, expected portfolio return, expected volatility, constraint checks, marginal risk contributions, and any infeasibility warnings. Do not silently relax constraints.

Context Variables

• asset_ids and current_weights
• expected_returns and covariance_matrix
• minimum_weights and maximum_weights
• sector_mapping and sector_caps
• turnover_limit, volatility_ceiling, and transaction_cost_assumptions

Expected Output Sample

Illustrative structure:

{
  "status": "feasible",
  "optimized_weights": [
    {"asset_id": "ASSET_A", "weight": null},
    {"asset_id": "ASSET_B", "weight": null}
  ],
  "portfolio_metrics": {
    "expected_return": null,
    "expected_volatility": null,
    "estimated_turnover": null
  },
  "constraint_checks": {
    "weights_sum_to_one": true,
    "asset_bounds_satisfied": true,
    "sector_caps_satisfied": true,
    "turnover_limit_satisfied": true
  },
  "marginal_risk_contributions": [],
  "warnings": []
}

Best Practice Tips

• Recalculate every portfolio metric outside the language model before approving an allocation.
• Require an explicit infeasibility result rather than allowing the model to modify a bound or constraint.
• Run sensitivity cases for expected returns, covariance estimates, turnover assumptions, and concentration limits.

Prompt 15 — VaR Backtesting and Exception Diagnostics

Production-ready Prompt Text:

Backtest the supplied daily Value at Risk forecasts against realized portfolio returns. Identify breach dates, calculate the exception count and exception rate, test for exception clustering using the diagnostics supplied by the user, and compare results with user-defined review thresholds. Return valid JSON containing summary_statistics, breaches, diagnostic_results, threshold_status, data_quality_warnings, and remediation_actions. Do not infer missing returns or forecasts.

Context Variables

• dates, realized_returns, and var_forecasts
• confidence_level and forecast_horizon
• diagnostic_method and decision_thresholds
• portfolio_changes and known data-quality events

Expected Output Sample

Illustrative structure:

{
  "summary_statistics": {
    "observation_count": null,
    "exception_count": null,
    "exception_rate": null
  },
  "breaches": [
    {"date": "YYYY-MM-DD", "realized_return": null, "var_forecast": null}
  ],
  "diagnostic_results": [],
  "threshold_status": "review_required",
  "data_quality_warnings": [],
  "remediation_actions": []
}

Best Practice Tips

• Align forecast and return timestamps before running the test.
• Investigate clustered breaches separately from isolated exceptions.
• Preserve the forecast series used at the time; do not replace it with a later recalculation.

Prompt 16 — Competitive Landscape and Industry Structure Map

Production-ready Prompt Text:

Using only the supplied company filings, industry reports, and market data, create a competitive landscape for the target company. Separate verified facts from analyst interpretation. Compare competitors across business model, customer segment, geographic exposure, pricing approach, distribution, cost drivers, and stated strategic priorities. Return valid JSON with market_structure, competitor_profiles, differentiation_claims, evidence_gaps, and research_questions. Attach a source ID and reporting date to every factual assertion.

Context Variables

• target_company and competitor_list
• source_documents with source IDs, publication dates, and reporting periods
• industry_scope and geographic_scope
• comparison_dimensions

Expected Output Format

Illustrative structure:

{
  "market_structure": {
    "verified_facts": [],
    "analyst_interpretations": []
  },
  "competitor_profiles": [
    {
      "company": "COMPANY_A",
      "comparison_points": [],
      "source_ids": []
    }
  ],
  "differentiation_claims": [],
  "evidence_gaps": [],
  "research_questions": []
}

Usage Notes

• Reject unsupported market-share, pricing, or growth claims instead of estimating them.
• Keep reporting periods aligned when comparing competitors.

Prompt 17 — Investment Thesis, Variant View, and Falsification Tests

Production-ready Prompt Text:

Build a structured investment thesis from the supplied evidence. State the base thesis, the market expectation provided by the analyst, the proposed variant view, supporting evidence, counterevidence, and explicit falsification tests. Distinguish facts, assumptions, and judgments. Return valid JSON with thesis, consensus_assumptions, variant_view, evidence_for, evidence_against, falsification_tests, monitoring_indicators, and unresolved_questions. Do not create a consensus view when none is supplied.

Context Variables

• company_name and investment_horizon
• verified_evidence and source IDs
• consensus_assumptions supplied by the analyst
• valuation_outputs and risk factors

Expected Output Format

Illustrative structure:

{
  "thesis": "",
  "consensus_assumptions": [],
  "variant_view": "",
  "evidence_for": [],
  "evidence_against": [],
  "falsification_tests": [
    {"test": "", "measurement": "", "review_date": "YYYY-MM-DD"}
  ],
  "monitoring_indicators": [],
  "unresolved_questions": []
}

Usage Notes

• Write falsification tests before drafting the recommendation.
• Use observable indicators and defined review dates rather than vague thesis statements.

Prompt 18 — Comparable-Company Screening and Multiple Reconciliation

Production-ready Prompt Text:

Evaluate the supplied comparable-company set. Check fiscal-period alignment, accounting definitions, currency treatment, enterprise-value inputs, and metric availability. Calculate requested valuation multiples only from provided values, summarize the distribution, and reconcile the target company's premium or discount to operating and financial differences. Return valid JSON with included_peers, excluded_peers, exclusion_reasons, calculated_multiples, distribution_summary, target_comparison, and data_quality_flags.

Context Variables

• target_company and proposed_peer_set
• market_values, net_debt, revenue, earnings, and cash-flow metrics
• reporting_periods, currencies, and accounting definitions
• peer inclusion and exclusion rules

Expected Output Format

Illustrative structure:

{
  "included_peers": [],
  "excluded_peers": [
    {"company": "", "reason": ""}
  ],
  "calculated_multiples": [],
  "distribution_summary": {},
  "target_comparison": {
    "premium_or_discount": null,
    "supported_drivers": [],
    "unsupported_hypotheses": []
  },
  "data_quality_flags": []
}

Usage Notes

• Exclude a peer when the required metric cannot be reconciled; do not fill the gap with an invented estimate.
• Retain both reported values and normalized values so reviewers can reproduce each multiple.

Prompt 19 — Catalyst Calendar and Evidence-Based Event Monitor

Production-ready Prompt Text:

Create a catalyst calendar from the supplied dated sources and analyst assumptions. For each event, record the date or date range, event type, source, thesis linkage, measurable confirmation signal, downside signal, and required follow-up. Mark dates as confirmed, estimated, or unknown. Return valid JSON with confirmed_events, estimated_events, unknown_date_events, monitoring_rules, and stale_source_warnings. Never convert an estimated date into a confirmed date.

Context Variables

• company_calendar and reporting_calendar
• management statements and regulatory milestones
• industry events and macroeconomic releases supplied by the analyst
• thesis drivers and risk indicators

Expected Output Format

Illustrative structure:

{
  "confirmed_events": [
    {
      "event": "",
      "date": "YYYY-MM-DD",
      "source_id": "",
      "thesis_linkage": "",
      "confirmation_signal": "",
      "downside_signal": ""
    }
  ],
  "estimated_events": [],
  "unknown_date_events": [],
  "monitoring_rules": [],
  "stale_source_warnings": []
}

Usage Notes

• Refresh the calendar when a company changes guidance or timing.
• Separate the expected business impact from the possible market reaction.

Prompt 20 — Source-Grounded Investment Memo with Red-Team Review

Production-ready Prompt Text:

Draft an investment memo using only the supplied research package. Include executive_summary, business_overview, thesis, valuation, catalysts, risks, counterarguments, falsification_tests, and open_questions. Cite a source ID for each factual statement and label every estimate or analyst judgment. Then perform a red-team review that identifies unsupported claims, stale evidence, inconsistent assumptions, omitted downside cases, and calculations requiring external verification. Return valid JSON only.

Context Variables

• research_package with source IDs and dates
• valuation_models and scenario_outputs
• portfolio_context and investment_horizon
• house_style and approval_requirements

Expected Output Format

Illustrative structure:

{
  "memo": {
    "executive_summary": "",
    "business_overview": "",
    "thesis": [],
    "valuation": {},
    "catalysts": [],
    "risks": [],
    "counterarguments": [],
    "falsification_tests": [],
    "open_questions": []
  },
  "red_team_review": {
    "unsupported_claims": [],
    "stale_evidence": [],
    "inconsistent_assumptions": [],
    "omitted_downside_cases": [],
    "calculations_to_verify": []
  }
}

Usage Notes

• Run the red-team pass separately from the drafting pass to reduce confirmation bias.
• Block publication when a material factual statement lacks a traceable source.