research-project-designer

What This Skill Does

The research-project-designer is a specialized computational science co-pilot designed to act as an uncompromising peer reviewer and system architect for drug discovery and biophysics workflows. It transitions the user from blind experimentation to physically grounded methodology by subjecting every proposal to a mandatory 'Fatal Flaw Audit' that evaluates dynamic target behavior, thermodynamic realities, data limitations, and current state-of-the-art obsolescence.

This skill is built for professionals who require rigorous analysis of molecular modeling, structural bioinformatics, and AI-driven drug discovery pipelines. It does not provide fluffy validation; instead, it enforces mathematical and physical rigor, mapping your objectives onto a Hierarchical Solution Matrix ranging from rapid, low-precision Python scripts to high-compute, high-precision Free Energy Perturbation (FEP) simulations.

Installation

Execute the following command in your terminal to integrate this agent skill into your workflow: clawhub install openclaw/skills/skills/chenbq18/research-project-designer

Use Cases

• Methodology Development: Designing an end-to-end pipeline for binding affinity prediction using AlphaFold structures.
• Code Review: Debugging complex NumPy/MDAnalysis/RDKit/OpenMM scripts for molecular dynamics (MD) trajectory analysis.
• Feasibility Auditing: Determining if a proposed binding pocket search using grid-based methods is physically sound or prone to 'static fallacy'.
• Workflow Optimization: Transitioning from coarse-grained docking to high-precision QM/MM or FEP workflows based on project-specific resource availability.

Example Prompts

1. "I'm trying to identify cryptic pockets on this kinase domain using a 50ns MD simulation and standard solvent probe clustering. Does this approach suffer from sampling bias?"
2. "Review my RDKit script for pharmacophore generation; I'm seeing an unexpected coordinate transformation error when parsing the PDB file."
3. "Should I use AutoDock Vina or transition to a deep learning-based scoring function for my library screening, given my current constraints on GPU resources?"

Tips & Limitations

• Physical Priority: The agent is programmed to prioritize physical law over software ease. Expect rigorous feedback on desolvation penalties and electrostatic mismatches.
• Strict Workflow: The agent forces a 5-phase execution workflow. If you want a quick fix without the audit, explicitly state: "Just fix the code, do not audit the methodology."
• Compute Awareness: Always define your compute budget (e.g., 'I have an A100 GPU for 24 hours') so the agent can provide the most relevant 'Plan' from the Hierarchical Solution Matrix.