ligandmpnn
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides high-level process instructions for ligand-aware protein design. No malicious patterns, obfuscation, or safety bypasses were identified.
- [NO_CODE]: This skill consists entirely of markdown-based workflow instructions. It does not include scripts, binaries, or references to third-party package managers.
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process external scientific data (protein structures and ligand files). While this represents a theoretical ingestion surface for data-driven influence, the instructions focus on technical verification and scientific plausibility. Evidence Chain: 1. Ingestion points: Input structures and ligand identities (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Implied command execution and file saving. 4. Sanitization: Absent.
Audit Metadata