solublempnn
Warn
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- Unverifiable Dependencies (MEDIUM): The skill instructs the user to clone and execute code from an untrusted GitHub repository (dauparas/ProteinMPNN). This source is not on the trusted organizations list, posing a risk of executing unverified external scripts.
- Indirect Prompt Injection (LOW): The skill processes untrusted PDB files which provides a surface for data-driven attacks if the underlying tool handles malicious metadata or structural information. Evidence Chain: 1. Ingestion points: Reads PDB files via '--pdb-path backbone.pdb' (SKILL.md). 2. Boundary markers: None; uses direct file path arguments. 3. Capability inventory: Executes local Python scripts ('protein_mpnn_run.py') and remote Modal functions. 4. Sanitization: No evidence of input validation or sanitization for processed PDB data.
- Command Execution (SAFE): Use of command-line tools and Python scripts is consistent with the primary purpose of protein sequence design.
Audit Metadata