Cloning Agent Skill Development

A computational molecular biology skill for AI agents that designs, verifies, and exports DNA cloning strategies using pydna.

---

📁 Repository Structure

cloning_agent/
├── SKILL.md                         # Main skill instructions (The actual skill)
├── references/                      
│   └── pydna_patterns.md            # Golden code patterns for pydna/OpenCloning
├── scripts/
│   ├── fetch_addgene.py             # Addgene sequence fetcher
│   ├── fetch_elabftw.py             # eLabFTW sequence fetcher
│   ├── search_addgene.py            # CLI tool to search Addgene
│   └── search_elabftw.py            # CLI tool to search eLabFTW
│
├── test_sequences/                  # Test GenBank and FASTA files
├── evals/                           # Testing & evaluation materials
├── cloning-agent-workspace/         # Test execution workspace
├── docs/                            # Project documentation
├── archive/                         # Old/experimental files
├── pyproject.toml                   # Python project config
└── uv.lock                          # Dependency lock file

---

🚀 Quick Start

Installing the Skill

This entire repository is the skill. To install it into your own agent workspace, simply clone this repository into your project's .agents/skills folder:

mkdir -p .agents/skills
git clone https://github.com/your-username/cloning-agent .agents/skills/cloning-agent

Using the Skill

Once installed, the skill will automatically trigger when you mention:

• DNA cloning, plasmid design, Gibson assembly, Golden Gate
• PCR primer design, restriction enzyme digestion
• GenBank files (.gb), pydna, OpenCloning JSON, Addgene/eLabFTW fetching

Example usage:

"I need to clone gene X from eLabFTW ('Storage Vector') into plasmid Y (Addgene 39296) using Gibson assembly"

Running Tests

If you cloned this repository just to develop/test the skill:

1. Quick validation (Test 6 - Simple ligation):

   cd cloning-agent-workspace/iteration-1/eval-6-simple-ligation/with_skill/outputs
   uv run python simple_ligation.py

2. See test results:

   cat cloning-agent-workspace/iteration-1/TEST_6_RESULTS.md

3. Run more tests:

   cat evals/TESTING_QUICKSTART.md  # Follow the guide

---

📊 Status

• Skill Version: v2 (Improved)
• Assessment Score: 72.5% → ~95% (after improvements)
• Tests Passed: 1/8 (Test 6: 100%)
• Ready for: Full test suite execution

Test Results

Test	Name	Status	Files Needed
6	Simple ligation	✅ 100%	None
1	Gibson assembly	⏳ Ready	test_sequences/*.gb
3	Primer design	⏳ Ready	test_sequences/*.gb
4	Golden Gate	⏳ Ready	test_sequences/*.fasta
7	Overlap PCR	⏳ Ready	test_sequences/*.gb
2	Restriction cloning	⚠️ Need addgene #39296
0	Complex workflow	⏳ Ready	(fetch from DB)
5	Multi-fragment	⏳ Ready	(mixed sources)

---

🔧 Key Improvements Applied

1. Enhanced Description (40% → 95%)

Added comprehensive trigger keywords for better skill activation.

2. Added "Why" Explanations

Replaced prescriptive "MUST" statements with reasoning:

• Why check AGENTS.md (biological parameter matching)
• Why get approval (experiments are expensive)
• Why use golden patterns (consistency)

3. Added Example Workflow

Shows complete interaction from user request to JSON export.

4. Added Troubleshooting Guide

Covers 4 common scenarios:

• No assembly products (incompatible overhangs)
• Files not found (where to look, how to fetch)
• Multiple products (how to choose)
• Dependency failures (fallback strategies)

5. Organized Reference File

Added table of contents and clear sections for quick lookup.

---

📖 Documentation

For Users

• AGENTS.md - Your cloning preferences (read by skill)
• evals/TESTING_QUICKSTART.md - How to run tests
• docs/DONE.md - Complete project summary

For Developers

• evals/skill_assessment.md - Comprehensive assessment (40-point rubric)
• evals/CHANGES_COMPARISON.md - Detailed before/after
• docs/PLAN.md - Original development plan
• .agents/skills/cloning-agent/SKILL.md - Actual skill code

---

🎯 Next Steps

1. Run remaining tests (1, 3, 4, 7)
2. Review and iterate based on results
3. Optimize description using skill-creator's run_loop
4. Package skill for distribution
5. Deploy to production

---

🔗 Quick Links

• Test sequences: test_sequences/ (9 files)
• Test cases: evals/evals.json (8 scenarios)
• Assessment: evals/skill_assessment.md
• Results: cloning-agent-workspace/iteration-1/
• Backup: .agents/skills/cloning-agent.backup/

---

📝 Notes

• Skill follows Anthropic skill-creator best practices
• Uses pydna for DNA assembly simulation
• Exports OpenCloning compatible JSON
• Test files are realistic GenBank and FASTA formats

---

Status: ✅ Assessment complete, improvements applied, first test passed!

Ready for: Full testing and production deployment