antibody-structure-prediction-tfold

Pass

Audited by Gen Agent Trust Hub on Mar 25, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches the tFold source code from the official TencentAI4S GitHub repository, which is a recognized scientific research entity.- [EXTERNAL_DOWNLOADS]: Downloads pre-trained model weights (ESM-PPI, AlphaFold, tFold-Ag) via library-managed functions such as tfold.model.esm_ppi_650m_ab().- [COMMAND_EXECUTION]: Uses standard shell commands (git clone, pip install) for environment setup and dependency installation from official package registries.- [COMMAND_EXECUTION]: Performs local file system operations to manage biological sequence data and invokes scientific processing scripts like generate_msa to generate input for the neural network.- [PROMPT_INJECTION]: The skill processes user-supplied protein sequences, which constitutes a surface for potential indirect prompt injection.
  • Ingestion points: The heavy_chain_sequence, light_chain_sequence, and antigen_sequence parameters in SKILL.md.
  • Boundary markers: No delimiters or explicit instructions are provided to isolate these inputs from the execution logic.
  • Capability inventory: The skill setup involves command-line tools (git, pip) and the code performs file writes (open().write()).
  • Sanitization: No explicit validation or escaping of the input sequences is detailed in the provided code snippets.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 25, 2026, 05:56 AM