The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates downloading the official research code and model checkpoints.
Evidence: Instructions point to the official GitHub repository at https://github.com/gcorso/DiffDock.git and the MIT CSAIL Docker image rbgcsail/diffdock.
Evidence: The skill mentions that model checkpoints (~500MB) are downloaded automatically during the first run, which is standard for pre-trained machine learning models.
[COMMAND_EXECUTION]: The skill relies on standard shell commands to execute its scientific workflows.
Evidence: Workflows use python -m inference to run the model and helper scripts like scripts/analyze_results.py for post-processing.
Evidence: Installation instructions use conda env create and docker run to establish the execution environment.
[SAFE]: The dynamic module loading (__import__) detected in scripts/setup_check.py is a routine method for environment verification.
Evidence: The function check_package uses __import__ to confirm that required dependencies like numpy, torch, and rdkit are installed before the user starts a docking job.

diffdock