esm

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill documentation guides the user to install the esm and flash-attn packages from standard package registries. These are well-known libraries in the bioinformatics and machine learning communities.
  • [COMMAND_EXECUTION]: Provides routine shell commands for package installation using uv pip. These commands are standard for setting up the required Python environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external biological data, including protein sequences and PDB structure files. This constitutes an ingestion point for untrusted data. However, the data is processed by specialized protein language models (ESM) which interpret the inputs as biological representations rather than natural language instructions, significantly limiting the potential for indirect prompt injection attacks.
  • [DATA_EXFILTRATION]: The skill connects to the official EvolutionaryScale Forge API (forge.evolutionaryscale.ai) for high-performance inference. This is a documented, legitimate use of a remote service for the skill's primary purpose. It handles user authentication via tokens in a standard manner.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 06:09 PM