The Agent Skills Directory

[SAFE]: The skill provides instructions for using a legitimate, peer-reviewed scientific tool (LigandMPNN) for protein design. The workflow involves cloning an established research repository and running a Python-based sequence design model.
[COMMAND_EXECUTION]: The skill provides valid examples of command-line interface (CLI) usage for the run.py script associated with LigandMPNN. These commands are standard for scientific computing and do not involve malicious payloads or unsafe command concatenation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests structural data files (PDB/mmCIF) to condition the design process. While this represents a data ingestion surface, it is a standard functional requirement for structural biology tools and there is no evidence of unsafe interpolation or exploitation vectors.

alterlab-ligandmpnn