Portfolio Stress Testing

## Role and objective
Stress test [PORTFOLIO] for an investor with [RISK TOLERANCE] risk tolerance.

## Analytical framework
1. Historical analogues: 2008 GFC, 2020 COVID crash, 2022 rate shock — estimate each holding's behaviour using its own history or sector proxies, stating the method.
2. Hypothetical shocks: rates +200bp, recession with earnings −30%, USD +15%, sector-specific shock for the largest exposure.
3. Separate security-specific risk, concentration, factor, and currency effects.
4. Check liquidity: how much could be sold within five days without major impact.

## Required output
- Scenario × estimated loss table
- Top loss contributors per scenario
- Whether losses fit the stated risk tolerance

## Professional report format
Deliver the analysis as a polished financial research report, not a chat reply. Use the language of the user's request. Apply this presentation standard without expanding beyond the research scope or overriding a requested word limit.

- Start with a concise report title, company or asset (if supplied), analysis date and data cut-off. If the cut-off is unknown, state that; do not invent it. Follow with an executive summary: conclusion, strongest evidence, principal risk and confidence with a reason.
- Use a clear hierarchy of numbered headings and short, well-written analytical paragraphs. Explain what each financial finding means rather than listing numbers without interpretation. Keep the tone measured, precise and institutional; avoid emojis, hype, decorative symbols and conversational introductions or sign-offs.
- Use compact tables for financial history, peer comparisons, forecasts or scenarios only where relevant and supported by data. Split wide tables into smaller ones suitable for a portrait A4 page. Label periods, currency, units and actual versus forecast figures clearly; distinguish reported facts, guidance, consensus and assumptions. Use consistent decimal precision and explain metric definitions. Mark missing values as unavailable, never as zero.
- Typography and visual identity: set the entire report — headings, body text and tables — in Times New Roman (or the closest available serif), the standard typeface of professional equity research. Use a black institutional theme: black for headings, rules, table headers and emphasis, mid-gray for supporting and secondary text, on a clean white page; no blue, no bright or decorative colours and no coloured page backgrounds.
- Figures are part of the report, not decoration: whenever the analysis discusses data — financial history, growth, margins, segment mix, comparisons, scenarios or sensitivities — include a picture of that data (a chart, diagram or map) next to the table whenever the assistant can create images: simple line, bar, stacked-area or waterfall charts that make the trend or comparison visible at a glance, in the same black, white and gray theme. Number figures in order (Figure 1, Figure 2, ...), refer to each one from the text, and give every figure a title, axis labels with units, the period covered and a dated source line beneath it. Plot only figures traceable to cited sources; draw forecasts, scenarios and illustrative values as dashed or shaded series and label them as such; keep charts simple enough to read at A4 width. If the assistant cannot create images, state that plainly and let the table carry the data — never present ASCII art or a text sketch as if it were a chart.
- Place traceable, dated source references next to material claims and table rows, with full source titles, dates and links in a final Sources and methodology section. Do not fabricate sources. Explain assumptions, calculation methods and unresolved data gaps; reconcile totals and distinguish percentage changes from percentage-point changes.
- Finish with a balanced conclusion answering the research question, key risks, contradictory evidence and the indicators that would change the conclusion. Include the analytical-exercise, not-investment-advice limitation.
- Make the output ready for document export: clean Markdown headings, restrained bold emphasis, readable paragraphs and no code fences around the report. Avoid ASCII art, oversized tables and empty decorative cover pages. For a longer report, use logically separated sections and a concise contents list only when useful.
- If the assistant supports creating documents, also provide an actual PDF using a classic institutional research layout: all text in Times New Roman serif, black headings, rules and table headers on a white page with gray supporting text, each figure embedded where it is discussed with its caption and source, A4 portrait pages, generous margins, repeated table headers and page numbers; avoid orphan headings and clipped tables. Otherwise provide the complete formatted report text and clearly state that no PDF file was generated. Never claim an attachment exists unless it was actually created.

## Evidence standard
- Separate clearly: (a) reported historical facts, (b) management guidance, (c) third-party/analyst estimates, (d) your own model assumptions. Label each.
- Cite dated, traceable sources for every material claim. Prefer company filings (10-K, 10-Q, 20-F, annual reports), investor-relations materials, regulatory disclosures, and reputable primary data.
- Do not invent figures, citations, or quotes. If a data point is unavailable or you cannot verify it, say so explicitly and state what would be needed.
- Include contradictory evidence and at least one credible alternative explanation.
- State when a method is unsuitable or incomplete for this situation.
- This is an analytical exercise, not investment advice. Keep the analysis separate from any decision.