---

name: generate-report description: Generate comprehensive markdown report with ranked candidates

Generate Report Skill

When to Use This Skill

Use as the final step to create a human-readable markdown report with all results, recommendations, and safety analysis.

Data Storage Structure

Reads from:

• data/{assembly}/genome_metadata.json - Cached genome metadata
• output/{run}/ranked_candidates.json - From score-rank step
• output/{run}/essential_genes.json - From identify-genes step
• output/{run}/literature_search.json - From literature search (optional)
• output/{run}/blast_results.json - From blast-screen step

Writes to:

• output/{run}/report.md - Final markdown report
• output/{run}/figures/summary_dashboard.png - Final dashboard

Instructions

Step 1: Load All Data Files

Use read_file for:

• output/{run}/ranked_candidates.json
• output/{run}/essential_genes.json
• output/{run}/literature_search.json (optional - if literature was searched)
• output/{run}/blast_results.json
• data/{assembly}/genome_metadata.json (cached input data)

Tip: If literature references are needed for the report but literature_search.json doesn't exist, use the literature-search utility skill now to enrich the report.

Step 2: Generate Report

Create output/{run}/report.md following the structure below.

Use the report template in references/report_template.md as a guide.

Step 3: Create Summary Dashboard

Generate a final summary visualization:

Copypython .deepagents/skills/generate-report/scripts/create_dashboard.py \
  --data-dir output/{run}/ \
  --output output/{run}/figures/summary_dashboard.png

This creates a multi-panel dashboard showing:

• Top 3 candidates with scores
• Safety statistics pie chart
• Gene essentiality scores
• Workflow completion checklist

Step 4: Present Final Results

Output this summary to the user:

Copy## Workflow Complete

**Report Generated:** `output/{run}/report.md`

**Top Recommendation:**
- **Candidate:** {top_candidate}
- **Target Gene:** {gene_name} - {gene_function}
- **Combined Score:** {score}
- **Safety Status:** Safe
- **Key Strength:** {rationale}

**Sequence (300bp):**

{sequence}

Copy
**Files Created:**
- `output/{run}/report.md`
- `output/{run}/figures/summary_dashboard.png`

**Dashboard:** [Show summary dashboard]

---
**Next Steps:**
1. Review the full report at `output/{run}/report.md`
2. Order synthesis of top candidate sequence
3. Design feeding bioassay protocol

Report Structure

1. Executive Summary

• Target species
• Number of candidates evaluated
• Top recommendation with justification

2. Top Candidates Table

3. Detailed Candidate Analysis

For top 3 candidates:

• Gene name and function
• Sequence (formatted as code block)
• GC content, length, position
• Off-target matches (human/honeybee)
• Safety status with icon

4. Off-Target Safety Analysis

• Table of all candidates with match lengths
• Summary statistics (safe/caution/reject counts)

5. Essential Genes Identified

• Table showing top genes evaluated
• Evidence sources (orthology, literature)

6. Literature References (if available)

• Papers found for this species
• Key genes mentioned
• (Note: If no literature search was run, state "No literature search performed")

7. Recommendations

• Which candidate to synthesize first
• Alternative candidates
• Testing protocol suggestions

8. Methods Summary

• Genome source (NCBI assembly ID)
• Essential genes database
• Off-target screening (BLAST parameters)
• Safety thresholds applied

Expected Output

All outputs go in output/{run}/:

• report.md - Professional markdown report
• figures/summary_dashboard.png - Final dashboard
• Complete audit trail of all intermediate files and plots

Available Tools

• read_file - Load JSON data
• write_file - Create report
• shell - Run dashboard generation script

Report Quality Checklist

Before presenting to user, verify:

• All sections populated
• Top candidate clearly identified
• Safety analysis complete
• Figures referenced correctly
• No placeholder text remaining
• Scientific accuracy of claims