Summary

Add autoresearch skill for autonomous, continuous research with a two-loop architecture.

What's New

• Inner Loop: Rapid experiment iteration with clear measurable outcomes
• Outer Loop: Periodic synthesis, pattern discovery, and direction setting
• Research Workspace Templates: State tracking, findings synthesis, decision log
• Example Project: LoRA rank convergence study demonstrating the workflow
• Configuration Options: Loop intervals, auto-commit, workspace settings

Use Cases

• Literature surveys with iterative hypothesis testing
• Benchmark optimization (e.g., "make val_loss go down")
• Mechanistic interpretability studies
• Hyperparameter discovery

Files Added

skills/research/autoresearch/
├── SKILL.md                           # Main skill documentation
├── templates/
│   ├── research-state.yaml            # State tracking template
│   ├── findings.md                    # Synthesis template
│   └── research-log.md                # Decision log template
└── examples/
    ├── README.md                      # Example overview
    └── lora-rank-study.md             # Example research project

Testing

• Skill YAML frontmatter validated
• Templates render correctly
• Example follows autoresearch methodology
• DESCRIPTION.md updated with skill overview

Related Skills

Integrates with: arxiv, research-paper-writing, web-search, notebooklm

Documentation

Full documentation in SKILL.md including:

• Two-loop architecture explanation
• Workspace structure
• Research discipline guidelines
• Configuration options
• Best practices

Inspired by: Andrej Karpathy's autoresearch methodology
Compatible with: agentskills.io open standard