Professional Research Report Generation

Overview

Generate structured, professional-grade financial research reports. Follow the conventions used by securities firms and asset managers, and output complete Markdown reports that can be used directly for investment research reference.

Saving and Delivering a PDF

When the user requests a PDF, send the complete Markdown report to write_file with a .pdf target (prefer artifacts/report.pdf under the active run directory). The tool renders a real, paginated PDF without requiring system Pango/Cairo. Check its returned status, path, and bytes_written; include the returned download_url as a Markdown link labelled with the report's filename. The Web chat downloads it with authentication and IM channels attach the generated PDF. A successful write is evidence that the returned path exists; another path's earlier rejection does not invalidate it. If rendering fails, report that tool error instead of claiming the PDF was created. Do not write HTML or base64 data as the content of a .pdf file.

Report Types and Structure

Type Classification

Standard Structure Template

# [Report Title]

> **Rating**: Buy | **Target Price**: ¥XX | **Current Price**: ¥XX
> **Analyst**: Vibe-Trading | **Date**: 2026-03-28

## Summary

3-5 key points, each 1-2 sentences, covering the core conclusion.
Use explicit language such as bullish / bearish / neutral. Do not be vague.

## Core Views

### View 1: [One-line heading]
2-3 paragraphs of discussion, supported by data.

### View 2: [One-line heading]
2-3 paragraphs of discussion, supported by data.

### View 3: [One-line heading]
2-3 paragraphs of discussion, supported by data.

## Main Analysis

### [Section 1: for example Industry analysis / Company fundamentals / Strategy logic]
Detailed analysis...

### [Section 2: for example Financial analysis / Competitive landscape / Backtest results]
Detailed analysis...

### [Section 3: for example Valuation analysis / Catalysts / Risk-reward]
Detailed analysis...

## Data Appendix

Key data tables...

## Risk Warnings

1. [Risk 1]: specific description + trigger condition + impact level
2. [Risk 2]: ...
3. [Risk 3]: ...

## Investment Recommendation

Clear action recommendation, including direction, position size, and time horizon.

---
*Disclaimer: This report is generated by an AI quantitative system for research reference only and does not constitute investment advice.*

Rating System

Equity Ratings

Strategy Ratings

Markdown Formatting Standards

Heading Hierarchy

# H1: Report title (only one)
## H2: Main sections (summary / views / body / risks / recommendation)
### H3: Subsections
#### H4: Maximum depth; use bold text instead of deeper nesting

Table Standards

| Metric | 2024A | 2025E | 2026E | Notes |
|------|-------|-------|-------|------|
| Revenue (100m RMB) | 500.2 | 580.0 | 660.0 | +15% YoY |
| Net profit (100m RMB) | 45.3 | 55.0 | 65.0 | +22% YoY |
| PE(TTM) | 25.0x | 20.5x | 17.3x | Valuation compressing |

Table rules:

• Right-align numbers where possible (when Markdown cannot do so, at least keep visual alignment clean)
• Put units in the header (100m / 10k / RMB)
• Forecast values use E (2025E), actual values use A (2024A)
• Keep 1 decimal place for percentages and 1-2 decimal places for monetary values

Number Formatting Standards

Chart Substitute (Text Description + Data Table)

Since Markdown output cannot embed images, use the following substitute:

### Equity Curve

The strategy equity started at 1.00 in 2020-01, peaked at 1.85 in 2021-02,
then suffered its largest drawdown from 2022.04 to 2022.10 (-25.3%), and
recovered by 2023.06 to a new high of 2.10.

| Time Point | Equity | Cumulative Return | Phase Event |
|---------|------|---------|---------|
| 2020-01 | 1.00 | 0% | Start |
| 2021-02 | 1.85 | +85% | Bull-market peak |
| 2022-10 | 1.38 | +38% | Drawdown low |
| 2023-06 | 2.10 | +110% | Recovery to new high |
| 2024-12 | 2.45 | +145% | Latest |

Professional Terminology Guide

Terms You Should Use (to improve professionalism)

Common Abbreviations

Backtest Report Template

After a backtest is completed, read the engine-written metrics.csv and equity.csv from that run. Declare reported measurements with the run directory or exact engine field as their grounding reference. A metrics file you write yourself is not evidence, and a previous run's archive cannot substantiate the current call. For list results, use the exact element reference (for example, call_id::data.positions[0].contribution_pct); correction hints only identify candidates to check and cite explicitly.

Use the verified output to generate:

# Backtest Report: [Strategy Name]

## Strategy Overview
- **Strategy type**: momentum / mean reversion / multi-factor / ...
- **Instruments**: 000001.SZ, 600519.SH, ...
- **Backtest period**: 2016-01-01 to 2026-01-01
- **Initial capital**: ¥1,000,000

## Performance Summary

| Metric | Strategy | Benchmark (CSI 300) | Excess |
|------|------|-------------|------|
| Cumulative return | +145.2% | +35.6% | +109.6% |
| Annualized return | 9.8% | 3.2% | +6.6% |
| Sharpe | 1.25 | 0.45 | - |
| Max drawdown | -18.5% | -32.4% | - |
| Calmar | 0.53 | 0.10 | - |
| Win rate | 58.3% | - | - |
| Trade count | 245 | - | - |

## Annual Performance

| Year | Return | Benchmark | Excess | Max Drawdown |
|------|------|------|------|---------|
| 2016 | +12.3% | -8.5% | +20.8% | -8.2% |
| ... | ... | ... | ... | ... |

## Risk Analysis
[Insert output from the `risk-analysis` skill]

## Improvement Suggestions
1. Specific suggestion 1...
2. Specific suggestion 2...

## Rating: Allocatable
Sharpe > 1.0 and max drawdown < 20%, indicating a good risk-reward profile.

Analysis Framework

Preparation Before Writing

1. Clarify the report type: stock / industry / strategy / backtest / flash note
2. Collect data: backtest results, fundamentals, industry data
3. Determine the core view: decide the conclusion first, then organize the supporting arguments
4. Choose the rating: based on quantitative metrics + qualitative judgment

Writing Principles

1. Put the conclusion first: the summary should contain the most important views, without too much buildup
2. Let data speak: every view should have at least one supporting data point
3. Keep logic clear: because A (data) → therefore B (judgment) → recommendation C (action)
4. Argue both sides: even in bullish reports, discuss risks; even in bearish reports, mention possible catalysts
5. Be actionable: recommendations should specify instruments / position sizing / time frame

Output Format

Output the complete Markdown research report directly, following the template structure above. Ensure that:

• it has a clear rating
• it is supported by quantitative data
• it includes at least 3 risk warnings
• it contains concrete action recommendations
• it ends with a disclaimer

Notes

1. Do not make overly precise forecasts: use ranges rather than single points (target price 25-30 RMB instead of target price 27.5 RMB)
2. Data consistency: all quoted numbers must be consistent throughout the report and never contradict each other
3. Mark timeliness: state both the data cut-off date and the report generation date
4. Disclaimer is mandatory: AI-generated reports must explicitly state that they are not investment advice
5. Avoid absolutes: "likely" is better than "certain", and "may" is better than "will"
6. China A-share specific risks: policy risk / delisting risk / goodwill impairment / major-shareholder selling / lock-up expiry
7. Crypto-specific risks: regulation / smart-contract exploits / liquidity risk / rug pulls