report-writing（报告写作）

Report Writing Skill

This skill provides structured guidance for transforming research findings into well-organized, professional reports. It ensures consistency, clarity, and completeness across all research outputs.

When to Use This Skill

Invoke this skill when:

• Finalizing Research: Converting raw research notes and findings into a formal deliverable
• Creating Documentation: Producing technical documentation, white papers, or analysis reports
• Synthesizing Multiple Sources: Combining insights from various sub-agent research tasks into a unified narrative
• Stakeholder Communication: Preparing reports for executive review, technical teams, or external audiences
• Knowledge Preservation: Documenting research methodology and findings for future reference

Do NOT use this skill for:

• Quick summaries or informal notes (use simple markdown instead)
• Real-time status updates (use TODO lists)
• Raw data dumps (use structured data files)

Report Structure Template

Every research report MUST follow this hierarchical structure. Adapt section depth based on report complexity.

1. Executive Summary

Purpose: Provide a standalone overview for readers who may not read the full report.

Contents:

• Research objective (1-2 sentences)
• Key findings (3-5 bullet points)
• Primary recommendations or conclusions
• Critical limitations or caveats

Length: 150-300 words (1 page maximum)

Writing Tip: Write this section LAST, after all other sections are complete.

2. Introduction/Background

Purpose: Establish context and frame the research question.

Contents:

• Problem statement or research question
• Why this research matters (business/technical impact)
• Scope boundaries (what IS and IS NOT covered)
• Brief overview of approach taken

Length: 200-500 words

3. Methodology

Purpose: Enable reproducibility and establish credibility.

Contents:

• Data sources consulted (with dates accessed)
• Search strategies and queries used
• Selection criteria for sources
• Tools and techniques employed
• Limitations of the methodology

Example Format:

### Data Collection
- Primary sources: [List with access dates]
- Search queries: [Exact queries used]
- Time range: [Date boundaries for research]

### Analysis Approach
- [Describe analytical framework]
- [Note any tools or models used]

4. Findings

Purpose: Present discovered facts objectively, without interpretation.

Contents:

• Organized by theme, source type, or chronology
• Each finding clearly attributed to source
• Quantitative data in tables/charts when applicable
• Direct quotes for critical evidence

Structure Options:

• Thematic: Group by topic or category
• Comparative: Side-by-side analysis of alternatives
• Chronological: Timeline of developments
• Source-based: Organized by information source

5. Analysis

Purpose: Interpret findings and extract meaning.

Contents:

• Patterns and trends identified
• Contradictions or gaps in evidence
• Implications of findings
• Comparison with existing knowledge
• Confidence levels for conclusions

Analysis Framework:

### Pattern Analysis
[What recurring themes emerge?]

### Gap Analysis
[What questions remain unanswered?]

### Confidence Assessment
- High confidence: [Findings with strong evidence]
- Medium confidence: [Findings with partial evidence]
- Low confidence: [Tentative findings requiring validation]

6. Conclusions

Purpose: Synthesize analysis into actionable insights.

Contents:

• Direct answers to research questions
• Prioritized recommendations (if applicable)
• Suggested next steps or future research
• Final assessment of confidence

Format:

### Key Conclusions
1. [Most important conclusion]
2. [Second conclusion]
3. [Third conclusion]

### Recommendations
1. [Priority 1 action item]
2. [Priority 2 action item]

### Future Research Directions
- [Unanswered questions to explore]

7. References

Purpose: Enable verification and further exploration.

Contents:

• All sources cited in the report
• URLs with access dates
• Proper attribution for all quoted material

Citation Formatting Guidelines

In-Text Citations

Use numbered references in square brackets for inline citations:

Recent studies indicate a 40% improvement in efficiency [1]. This aligns with
earlier findings on automation benefits [2, 3].

For direct quotes, include page numbers or section identifiers:

According to the official documentation, "the system supports up to 10,000
concurrent connections" [4, Section 3.2].

Reference List Format

Use a consistent format for all references:

Web Sources:

[1] Author/Organization. "Article Title." Website Name. URL. Accessed: YYYY-MM-DD.

Academic Papers:

[2] Author(s). "Paper Title." Journal/Conference Name, Year. DOI/URL.

Documentation:

[3] "Document Title." Product Name Documentation, Version X.X. URL. Accessed: YYYY-MM-DD.

News Articles:

[4] Author. "Headline." Publication Name, Date Published. URL.

Citation Best Practices

1. Always include access dates for web sources (content may change)
2. Prefer primary sources over secondary reports
3. Note version numbers for software documentation
4. Archive volatile sources when possible (use archive.org links)
5. Verify link validity before finalizing report

Writing Style Recommendations

Style Selection Guide

Executive Style

Characteristics:

• Lead with conclusions and recommendations
• Use bullet points liberally
• Limit technical jargon; define necessary terms
• Focus on business impact and ROI
• Keep paragraphs short (3-4 sentences max)

Example:

## Key Finding: Cloud Migration Reduces Costs by 35%

**Bottom Line**: Migrating to cloud infrastructure will reduce operational
costs by $2.4M annually while improving system reliability.

**Recommended Action**: Approve Phase 1 migration by Q2 2025.

**Risk Level**: Low - Similar migrations have 94% success rate.

Technical Style

Characteristics:

• Include implementation details
• Use precise technical terminology
• Provide code samples, configurations, or specifications
• Document edge cases and limitations
• Include performance metrics and benchmarks

Example:

## Implementation: Rate Limiting Configuration

The API gateway implements token bucket rate limiting with the following
parameters:

| Parameter | Value | Rationale |
|-----------|-------|-----------|
| Bucket Size | 1000 | Handles burst traffic |
| Refill Rate | 100/sec | Sustainable throughput |
| Key Strategy | IP + User ID | Prevents abuse while supporting legitimate use |

rate_limiter = TokenBucket( capacity=1000, refill_rate=100, key_func=lambda req: f"{req.ip}:{req.user_id}" )

Academic Style

Characteristics:

• Formal third-person voice
• Extensive literature review
• Detailed methodology documentation
• Statistical rigor where applicable
• Acknowledge limitations explicitly

Example:

## Literature Review

Previous research in automated code review systems has demonstrated
significant potential for defect detection. Smith et al. (2023) reported
a 23% reduction in production defects when implementing static analysis
tools [1]. However, Johnson and Lee (2024) note that these gains are
contingent upon proper configuration and team adoption [2].

The present study extends this work by examining the integration of
large language models into the review pipeline, an approach not
addressed in prior literature.

General Guidelines (All Styles)

1. Active voice preferred: "The team implemented" not "It was implemented by the team"
2. Specific over vague: "37% increase" not "significant increase"
3. Present tense for findings: "The data shows" not "The data showed"
4. Consistent terminology: Choose one term and use it throughout
5. Avoid hedging excess: Limit "may," "might," "could possibly"

Quality Checklist Before Submission

Structure Verification

• All seven standard sections present (or justified omission noted)
• Executive summary can stand alone
• Logical flow from introduction to conclusions
• Section lengths appropriate to content importance
• Headers and subheaders create clear hierarchy

Content Quality

• Research question clearly stated and answered
• All claims supported by cited evidence
• Findings and analysis clearly separated
• Contradictory evidence acknowledged
• Confidence levels stated for conclusions
• Limitations explicitly documented

Citation Integrity

• All sources cited in reference list
• All references cited in text
• URLs verified as accessible
• Access dates included for web sources
• No broken or placeholder citations

Writing Quality

• Consistent writing style throughout
• Technical terms defined on first use
• No unexplained acronyms
• Spell-check completed
• Grammar review completed
• Sentence length varied (not all long or all short)

Formatting

• Consistent heading styles
• Tables and figures numbered and titled
• Code blocks properly formatted
• Bullet points parallel in structure
• Page breaks at logical points (if applicable)

Final Review

• Report answers the original research question
• Recommendations are actionable
• Nothing critical missing from scope
• Appropriate length for audience and purpose
• Ready for intended audience

Examples of Report Sections

Example: Executive Summary

## Executive Summary

This report evaluates three cloud database solutions for the customer
analytics platform migration: AWS Aurora, Google Cloud Spanner, and
Azure Cosmos DB.

**Key Findings**:
- AWS Aurora offers the lowest total cost of ownership ($145K/year)
- Google Cloud Spanner provides superior global consistency guarantees
- Azure Cosmos DB integrates best with existing Microsoft infrastructure
- All three solutions meet performance requirements (< 50ms p99 latency)

**Recommendation**: Proceed with AWS Aurora for Phase 1, with architecture
designed to allow future multi-cloud expansion.

**Timeline**: Implementation achievable within Q2 2025 with existing team.

**Confidence Level**: High - Based on proof-of-concept testing and vendor
consultations.

Example: Methodology Section

## Methodology

### Research Approach

This analysis employed a mixed-methods approach combining:
1. Vendor documentation review
2. Technical proof-of-concept testing
3. Industry analyst report analysis
4. Peer organization interviews

### Data Sources

| Source Type | Sources Consulted | Date Range |
|-------------|-------------------|------------|
| Vendor Docs | AWS, GCP, Azure official documentation | Dec 2024 |
| Analyst Reports | Gartner, Forrester database evaluations | 2024 |
| Technical Tests | Internal POC environment | Dec 15-22, 2024 |
| Interviews | 3 peer organizations (anonymized) | Dec 2024 |

### Evaluation Criteria

Solutions were evaluated against weighted criteria:
- Performance (30%): Latency, throughput, scalability
- Cost (25%): TCO over 3 years including migration
- Reliability (20%): SLA guarantees, disaster recovery
- Integration (15%): Compatibility with existing stack
- Vendor Support (10%): Documentation, support quality

### Limitations

- POC testing limited to 72-hour duration
- Cost projections based on current pricing (subject to change)
- Interview sample size limits generalizability

Example: Findings Section

## Findings

### Performance Comparison

All three solutions demonstrated acceptable performance for the target
workload of 10,000 queries per second:

| Solution | Avg Latency | P99 Latency | Max Throughput |
|----------|-------------|-------------|----------------|
| AWS Aurora | 12ms | 45ms | 15,000 QPS |
| Cloud Spanner | 15ms | 42ms | 18,000 QPS |
| Cosmos DB | 18ms | 48ms | 12,000 QPS |

*Source: Internal POC testing, December 2024 [1]*

### Cost Analysis

Three-year total cost of ownership analysis:

**AWS Aurora**: $435,000
- Compute: $180,000
- Storage: $95,000
- Data transfer: $85,000
- Support: $75,000

**Google Cloud Spanner**: $520,000
- [Detailed breakdown...]

**Azure Cosmos DB**: $485,000
- [Detailed breakdown...]

*Source: Vendor pricing calculators and enterprise discount estimates [2, 3, 4]*

Example: Analysis Section

## Analysis

### Cost-Performance Trade-offs

While AWS Aurora offers the lowest TCO, Cloud Spanner's 20% higher cost
delivers measurably better global consistency. For applications requiring
strong consistency across regions, this premium may be justified.

The cost difference primarily stems from:
1. Cloud Spanner's TrueTime infrastructure overhead
2. AWS Aurora's more aggressive reserved instance discounts
3. Different approaches to cross-region replication

### Risk Assessment

| Risk | Likelihood | Impact | Mitigation |
|------|------------|--------|------------|
| Vendor lock-in | High | Medium | Abstract data layer |
| Price increases | Medium | Medium | 3-year commitment |
| Service outage | Low | High | Multi-region deployment |

### Confidence Assessment

**High Confidence**:
- Performance meets requirements (validated via POC)
- AWS Aurora is most cost-effective option

**Medium Confidence**:
- 3-year cost projections (pricing may change)
- Integration complexity estimates

**Low Confidence**:
- Long-term vendor roadmap alignment

Example: References Section

## References

[1] Internal Engineering Team. "Database POC Test Results." Internal
    Documentation. December 22, 2024.

[2] Amazon Web Services. "Amazon Aurora Pricing." AWS Documentation.
    https://aws.amazon.com/aurora/pricing/. Accessed: December 20, 2024.

[3] Google Cloud. "Cloud Spanner Pricing." Google Cloud Documentation.
    https://cloud.google.com/spanner/pricing. Accessed: December 20, 2024.

[4] Microsoft Azure. "Azure Cosmos DB Pricing." Azure Documentation.
    https://azure.microsoft.com/pricing/details/cosmos-db/.
    Accessed: December 20, 2024.

[5] Gartner. "Magic Quadrant for Cloud Database Management Systems."
    Gartner Research, November 2024. (Subscription required)

[6] Smith, J. and Chen, L. "Comparative Analysis of Distributed Databases."
    Proceedings of VLDB 2024. DOI: 10.14778/example.

Integration with Research Workflow

This skill integrates with the broader research workflow as follows:

Research Request → Data Collection → Analysis → [REPORT WRITING] → Verification → Delivery
                                                      ↑
                                                 This Skill

Inputs Expected:

• Completed research findings (from sub-agents or direct research)
• Original research request/questions
• Source materials and citations
• Any constraints (length, audience, format)

Outputs Produced:

• Formatted report following structure template
• Complete reference list
• Executive summary for quick consumption

Quality Gates:

• Report must pass quality checklist before marking complete
• All citations must be verifiable
• Conclusions must trace back to evidence in findings