A specialized Agent Skill designed for AI assistants (Gemini CLI, Claude Code, OpenCode) to autonomously acquire, analyze, and synthesize methodology frameworks from State-of-the-Art (SOTA) research.
Goal: Transform the literature review process from "reading papers" to "generating actionable research frameworks."
This is a self-contained skill project that agents can use to:
- Search academic databases (OpenAlex, Semantic Scholar) for solutions to specific engineering problems.
- Curate results using LLM reasoning to identify the most methodologically relevant papers.
- Extract concrete algorithms, code patterns, and validation strategies from PDFs.
- Synthesize a coherent "Methodology Section" draft and a practical "Implementation Guide."
This repository follows the Agent Skill standard:
.
├── SKILL.md # Skill definition and instructions for the Agent
├── scripts/ # Executable tools
│ ├── pipeline.py # Main orchestrator
│ ├── search.py # Search & Curation tool
│ ├── download.py # PDF Download tool
│ └── analyze.py # Extraction & Synthesis tool
├── lib/ # Core logic library
├── examples/ # Configuration templates
└── requirements.txt # Python dependencies
This repository is designed to be "discovered" by your agent. Point your agent to this directory, and it will recognize the capabilities defined in SKILL.md.
Prompt Example:
"Use the SOTA-Seeker skill to design a methodology for 'Low-resource LLM Fine-tuning'. create a config file first."
-
Install Dependencies:
pip install -r requirements.txt
-
Configure: Define your research problems in a YAML file:
cp examples/engineering_problems.yaml my_research.yaml # Edit my_research.yaml -
Run:
python scripts/pipeline.py --config my_research.yaml
- Methodology Draft (
TECHNICAL_REPORT.md): A synthesized, academically-toned text ready for your paper's methodology section. - Implementation Plan (
IMPLEMENTATION_GUIDE.md): A checklist-style guide for engineers to build the system.
MIT