experiment-report

name	experiment-report
description	This skill should be used when generating integrated reports from lab notebooks or improving existing reports. Triggered by requests like "generate report", "create summary", "refine report", "improve report", or "レポートを作成". Handles both initial generation (mechanical) and iterative refinement (AI-guided).

Experiment Report Management

Overview

Provides comprehensive report generation and refinement capabilities. Following the skill-creator pattern, this plugin handles both:

1. Init (Generation): Mechanical extraction and templating from lab notebooks
2. Refine (Improvement): AI-guided improvement for scientific rigor and clarity

Core Capabilities

1. Report Generation

Create integrated reports from completed lab notebooks using init_report.py.

Execution context: All commands assume execution from project root (where STEERING.md is located).

When to use: When ready to synthesize multiple experiments into a cohesive report.

Pre-Generation Checklist (MANDATORY)

Before running init_report.py, verify these requirements:

Step 1: Notebook Completeness

• All included notebooks have completed Results sections
• All included notebooks have completed Discussion sections
• All figures saved to results/exp##/ directory

Step 2: Project Alignment

• Report scope aligns with STEERING.md objectives
• Primary hypothesis/research question identified
• Notebooks collectively address the research question

Step 3: Evidence Verification

• List all figures needed for each finding
• Verify figure paths exist: ls ../results/exp##/
• Identify key statistics for each claim
• Note any missing evidence to address

Pre-Generation Questions (ask user):

1. "Which experiments are included?" (list Exp## numbers)
2. "Is the Discussion section complete in each notebook?"
3. "What is the main research question this report addresses?"
4. "What are the 2-3 key claims you want to make?"
5. "Are all figures saved to results/exp##/ directories?"

Workflow

1. Complete pre-generation checklist above
2. Run init script (from project root):

   python path/to/init_report.py --labnote notebook/labnote/Exp*.ipynb --output notebook/report/
   

3. Script generates template with claim-evidence structure
4. Fill in evidence tables for each finding (verify paths exist)
5. Complete quality gate checklists in each section
6. Output: notebook/report/Report_[title].md

Mapping rules (from lab notebooks to report):

Lab Notebook Section	Report Section	Transformation
Hypothesis + Background	Background	Synthesize context
Materials & Methods	Methods Summary	Consolidate procedures
Results (observations)	Findings	Structure by theme
Discussion (interpretation)	Synthesis	Integrate interpretations
Limitations	Limitations	Consolidate caveats
Next Steps	Future Directions	Prioritize follow-ups
Key conclusion	Executive Summary	Distill to 3-5 sentences

Command: /research-report

2. Report Refinement

Improve existing reports using references/refinement-guide.md criteria.

When to use: After initial report generation or when report needs improvement.

Workflow:

1. User specifies existing report to refine
2. Read current report content
3. Evaluate against refinement criteria:
   • Structure: Logical flow, clear sections
   • Scientific rigor: Fact/interpretation separation, evidence-based claims
   • Clarity: Concise writing, proper terminology
4. Provide specific improvement suggestions
5. Implement improvements (with user approval)

Refinement dimensions:

1. Structure & Organization:

   • Executive Summary: 3-5 sentences, captures essence
   • Logical flow: Each section builds on previous
   • Redundancy: No unnecessary repetition
   • Completeness: All findings addressed

2. Scientific Accuracy:

   • Facts vs interpretation: Clearly separated
   • Evidence: Every claim supported by data
   • Limitations: Acknowledged appropriately
   • Alternatives: Considered and discussed
   • Statistics: Properly reported

3. Writing Quality:

   • Clarity: Precise, unambiguous language
   • Conciseness: Efficient communication
   • Terminology: Consistent and appropriate
   • Figures: Properly referenced
   • Citations: Accurate and complete

Command: /research-refine

3. Report Structure

Standard report structure:

# [Report Title]

**Date**: YYYY-MM-DD
**Report Type**: Integrated Analysis Report

---

## Executive Summary

[3-5 sentences capturing key findings and implications]

## Background

[Research question, context, and rationale]

## Materials and Methods

[Consolidated methods from experiments]

## Findings

### Finding 1: [Title]
- **Observation**: [Factual description]
- **Evidence**: [Lab notebook references, figures]

### Finding 2: [Title]
[Repeat structure]

## Synthesis

[Integrated interpretation of findings]

## Limitations

[Acknowledged constraints and caveats]

## Future Directions

[Prioritized next steps]

## Conclusion

[Final synthesis and key takeaways]

## References

[Citations]

---

## Appendix

### Lab Notebooks
- Exp01_[name].ipynb
- Exp02_[name].md

### Supplementary Figures
[Links]

4. Quality Standards

Reports maintain separation between facts and interpretation (from research-project quality-standards.md):

In Findings section (Level 1: Facts):

• Present observations directly
• Reference source notebooks
• Include exact measurements
• Avoid interpretation

In Synthesis section (Level 2: Interpretation):

• Connect observations to meaning
• Cite supporting evidence
• Acknowledge assumptions
• Consider alternatives

In Conclusion section (Level 3: Broader implications):

• Synthesize evidence
• State confidence appropriately
• Suggest applications
• Identify unknowns

5. Figure Integration

Integrate figures from lab notebooks and analysis outputs into reports.

Figure sources:

• Jupyter notebooks: Inline outputs saved to results/exp##/
• Command-line tools: Output images in results/exp##/
• External analysis: Imported figures with proper attribution

Directory structure:

notebook/
├── labnote/
│   ├── Exp01_analysis.ipynb
│   └── Exp02_validation.md
├── report/
│   └── Report_Exp01-02_integrated.md
└── results/
    ├── exp01/
    │   ├── fig01_heatmap.png
    │   └── fig02_volcano.png
    └── exp02/
        └── fig01_validation.png

Markdown figure syntax:

![Figure 1: Heatmap of differential expression](../results/exp01/fig01_heatmap.png)

*Figure 1: Heatmap showing top 50 differentially expressed genes (Exp01).*

Figure naming convention:

• Format: fig##_description.{png,pdf,svg}
• Examples: fig01_heatmap.png, fig02_volcano_plot.pdf
• Use descriptive names for clarity

Best practices:

1. Save all figures to results/exp##/ during experiment execution
2. Use relative paths from report location (../results/...)
3. Include figure captions with experiment reference
4. Prefer PNG for raster, PDF/SVG for vector graphics
5. Number figures sequentially within each experiment

Extracting figures from Jupyter notebooks:

# In notebook cell, save figure explicitly
import matplotlib.pyplot as plt
fig.savefig('../results/exp01/fig01_heatmap.png', dpi=150, bbox_inches='tight')

6. PDF Export

Export final reports to PDF using pandoc with typst engine.

Prerequisites:

• pandoc (document converter)
• typst (PDF engine)

Basic export command:

pandoc Report_Exp01-02_integrated.md -o report.pdf --pdf-engine=typst

With typst template:

pandoc Report_Exp01-02_integrated.md -o report.pdf \
  --pdf-engine=typst \
  --template=assets/templates/report.typ

Template location: assets/templates/report.typ

Export workflow:

1. Complete and refine Markdown report
2. Verify all figure paths are correct (relative paths)
3. Run pandoc command from notebook/report/ directory
4. Review PDF output for formatting issues
5. Iterate if needed

Troubleshooting:

• Missing figures: Check relative paths from report directory
• Font issues: Ensure typst has access to required fonts
• Long tables: Consider splitting or using landscape orientation

Workflow Patterns

Pattern 1: Single Experiment Report

Scenario: Document one completed experiment

Steps:

1. Ensure lab notebook complete
2. Run: /research-report specifying single notebook
3. Review generated report
4. Refine: /research-refine to improve clarity

Use case: Individual experiment worthy of formal documentation

Pattern 2: Integrated Multi-Experiment Report

Scenario: Synthesize multiple related experiments

Steps:

1. Identify notebooks to include (e.g., Exp01-Exp05)
2. Run: /research-report specifying all notebooks
3. Script extracts and consolidates content
4. AI synthesizes findings across experiments
5. Refine: /research-refine to improve integration

Use case: Project milestone, manuscript preparation

Pattern 3: Iterative Refinement

Scenario: Improve existing report quality

Steps:

1. Identify report needing improvement
2. Run: /research-refine on existing report
3. Review suggestions organized by category:
   • Structure improvements
   • Scientific rigor issues
   • Writing clarity enhancements
4. Approve and implement changes
5. Iterate if needed

Use case: Preparing for presentation, submission, or review

Resources

scripts/

• init_report.py: Report generation script (executable)

commands/

• research-report.md: Report generation command (/research-report)
• research-refine.md: Report refinement command (/research-refine)

references/

• mapping-rules.md: Lab notebook → report section mapping
• refinement-guide.md: Quality criteria for report improvement

Usage Notes

Generation Best Practices

1. Timing: Generate reports when:

   • Multiple related experiments complete
   • Ready for project milestone
   • Preparing manuscript
   • Need formal documentation

2. Scope: Include notebooks that:

   • Address related questions
   • Build on each other
   • Collectively tell a story

3. Preparation: Before generating:

   • Ensure all lab notebooks complete
   • Review notebook quality
   • Identify key findings

Refinement Best Practices

1. Multiple passes: Refine in stages:

   • First pass: Structure and organization
   • Second pass: Scientific rigor
   • Third pass: Writing quality

2. Specific feedback: Request targeted improvements:

   • "Improve executive summary"
   • "Check fact/interpretation separation"
   • "Enhance clarity in Methods"

3. Version control: Commit before and after refinement

Common Improvements

Executive Summary:

• ❌ Too long (>5 sentences) or too vague
• ✅ Concise (3-5 sentences), captures essence

Findings:

• ❌ Mixed facts and interpretation
• ✅ Pure observations, references to notebooks

Synthesis:

• ❌ Unsupported claims, circular reasoning
• ✅ Evidence-based, acknowledges limitations

Limitations:

• ❌ Missing or too apologetic
• ✅ Honest, constructive, identifies solutions

Future Directions:

• ❌ Vague ("more research needed")
• ✅ Specific next experiments with rationale

Integration with Workflow

Typical flow:

1. Complete experiments → Lab notebooks (lab-notebook)
2. Refine hypotheses → Hypothesis validation (hypothesis-driven)
3. Generate report → Synthesis (experiment-report: init)
4. Improve quality → Refinement (experiment-report: refine)
5. Update project → STEERING.md (research-project)

Report triggers phase transitions:

• Generating report often signals end of Execution phase
• Moving to Integration or Publication phase
• Update STEERING.md accordingly

Examples

Example 1: Quick Single-Experiment Report

User: "Create a report for Exp03"